Pith. sign in

REVIEW 3 major objections 6 minor 180 references

How To Think About End-To-End Encryption and AI: Training, Processing, Disclosure, and Consent

T0 review · 3 major / 6 minor · reviewed 2026-08-10 · deepseek-v4-flash

Pith's one-line read The paper argues that training a shared AI model on end-to-end encrypted content is definitionally incompatible with E2EE, and only local or strictly per-user inference can preserve the guarantee.

desk verdict A genuinely useful framework for the E2EE/AI question, but Recommendation 1 overclaims by calling the incompatibility definitional when the paper's own caveat-feature category already relaxes the derivative definition. read the letter →

arxiv 2412.20231 v2 pith:RZ6HZP7Q submitted 2024-12-28 cs.CR cs.AI

classification cs.CRcs.AI
keywords end-to-endencryptionAIassistantsconfidentialitytrainingdataleakagelargelanguagemodelsprivacy-preservingmachinelearninguserconsentconsumerprotection
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

End-to-end encryption promises more than hiding message bytes: it promises that no one outside the sender and recipient can learn anything about a message's content, including any function of it. The paper argues that a shared AI model trained on end-to-end encrypted content is such a function, so letting other users query the model violates E2EE confidentiality by design. It distinguishes this from inference: processing encrypted content for AI is compatible with E2EE only when it happens on the user's own device, or when no third party can see or use the content and each user's content is used only for that user's request. The paper then argues that providers that ship AI features which use encrypted content must not market themselves as E2EE without qualification, and that AI features should be off by default and opt-in. The stakes are practical: Apple Intelligence, Samsung Galaxy AI, and Meta AI in WhatsApp are already deploying such integrations.

What carries the argument

The load-bearing object is the paper's definition of E2EE content, which includes any derivative of data the provider holds out as end-to-end encrypted; this definition is what makes the trained model itself E2EE content. The analytic device is a four-consideration framework: where the model runs, whether non-endpoint-local processing is privacy-enhanced, what type of confidentiality the enhancement provides, and whether the model is shared or per-user. A shared model fails the fourth consideration no matter what, an endpoint-local model passes all considerations, TEE-based processing supplies a different kind of security than E2EE, and per-user models can be compatible if the user's data is strictly isolated and statelessly processed.

What would settle it

The claim would be refuted by a shared model trained on E2EE content with a formal, information-theoretic guarantee that no output and no parameter depends on any user's message content; it would also be dissolved if a court or standards body explicitly defined E2EE confidentiality to cover only direct access to plaintext rather than functions of it.

Watch

Extended reading notes

Core claim

The paper's central claim is that using end-to-end encrypted content to train a shared AI model is definitionally incompatible with E2EE. E2EE confidentiality is defined to cover any derivative of message content, and a model trained on messages is a derivative of those messages; any user who can query the shared model can receive outputs that depend on other users' private messages. The paper argues that privacy-preserving training techniques, including differential privacy, data sanitization, federated learning, and multi-party computation, offer a spectrum of privacy rather than the binary guarantee of E2EE, so none of them can make shared-model training compatible. The same reasoning yields a four-part evaluation framework for AI assistants and a five-category taxonomy of applications ranging from strict E2EE to no E2EE.

Load-bearing premise

The entire argument rests on the definitional choice that E2EE confidentiality forbids any third party from learning any function of message content, and that a trained model is such a function; if a narrower definition of confidentiality is adopted, the claim that shared-model training is definitionally incompatible with E2EE does not follow.

Editorial extensions

If this is right

  • If the central claim is accepted, no shared model trained on E2EE content, no matter what privacy technique is used, can coexist with an honest claim that the service still provides E2EE.
  • Inference on E2EE content is compatible with E2EE only when it happens entirely on the user's device, or when the content is hidden from every third party and used exclusively to answer that user's request.
  • A messaging app that adds a cloud AI feature that is on by default and cannot be turned off is demoted, under the paper's taxonomy, to the no-E2EE category, even if its core messaging remains encrypted.
  • Providers that route E2EE content through AI features must qualify their E2EE marketing, or risk being deceptive under U.S. consumer-protection precedent.
  • AI features in E2EE systems should be off by default and activated only through explicit opt-in consent, with opt-out as easy as opt-in.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The derivative argument extends beyond training: any server-side computation on E2EE plaintext, including client-side scanning or lawful-access designs, produces outputs that are derivatives of E2EE content; the paper gestures at this in its Crypto Wars discussion but does not fully develop it.
  • A testable extension would be to attempt to build a shared model whose outputs are information-theoretically independent of the E2EE content it was trained on; the paper's definitional claim predicts this is impossible, so a construction would force a redraw of the derivative definition.
  • The taxonomy could be used as a pre-launch checklist: classify any proposed AI feature against the five categories, and let the category determine what defaults and marketing language are permissible.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 6 minor

Summary. This paper asks whether integrating AI assistants into end-to-end encrypted messaging systems, and training models on E2EE content, is compatible with the security guarantees of E2EE. It contributes a nine-term vocabulary and a five-category taxonomy (strict E2EE to no E2EE), a four-consideration technical evaluation framework (model location, privacy enhancement, guarantee type, shared vs. per-user), a survey of relevant US and EU legal frameworks and consent theory, and four numbered recommendations. Its headline normative claim is Recommendation 1: using E2EE content to train shared AI models is not compatible with E2EE. The paper applies its framework to Apple Intelligence, Samsung Galaxy AI, and Meta AI in WhatsApp, and is notably transparent about what vendor documentation does not disclose.

Significance. If the central claim is accepted, the paper provides a useful shared vocabulary and evaluation template for an urgent policy debate, and its legal synthesis is a valuable starting point for regulators. The paper's strengths include the explicit taxonomy, the structured evaluation framework with per-configuration verdicts, the candid "What We Still Don't Know" lists, and the willingness to identify open legal questions rather than overclaim certainty. It is a position/analysis paper rather than a formal security proof, and its central categorical recommendation needs refinement before the paper's claims can be fully endorsed; with that refinement, the paper would be a significant contribution to the cs.CR and security-policy literature.

major comments (3)
  1. [§3.1, §3.2, §7.1] Recommendation 1 is not entailed by the paper's own definitions. Section 3.1 defines E2EE content to include "any derivatives" of protected data (footnote 32), and Section 7.1 concludes that training a shared model on E2EE content "definitionally undermines" E2EE because the model is a derivative of its training data. However, Section 3.2 already grants Category 2 ("E2EE in current practice") to applications that expose plaintext-dependent content to third parties through caveat features such as link previews, and Section 2.1.5 concedes that link previewing "technically violates E2EE confidentiality." A URL in a message is primary content, and the preview service's output is a derivative of that content; under the paper's own definition, a link preview is exactly the kind of disclosure that Recommendation 1 declares categorically incompatible. The only stated distinction is that caveat features are "strictly limited" while AI processing involves "significant volumes" (§4.1.2), and footnote 34 explicitly declines to define a threshold. Recommendation 1 therefore requires either a principled threshold separating caveat features from model training or an explicit restriction to "strict E2EE" (Category 1); the paper should supply one of these before the central claim can be evaluated.
  2. [§4.2, Consideration #1, Consideration #4, Table 1] The opening of Section 4.2 states that "existing privacy-enhancing techniques cannot address the security concerns we raise around integrating AI assistants with E2EE applications," but the subsequent analysis concludes that endpoint-local models are "fully compatible with E2EE" (Consideration #1) and that per-user models "can be compatible with E2EE" (Consideration #4), and Table 1 marks the on-device configuration with a green tag. The summary sentence is therefore overbroad and contradicts the paper's own framework. It should be qualified to refer to configurations that involve third-party or shared processing, or the compatible configurations should be explicitly acknowledged as exceptions. As written, the section's framing obscures the paper's main message and reduces the reader's ability to rely on the framework summary.
  3. [§7.2, Recommendation 4] Recommendation 4 is in tension with Recommendation 1. Recommendation 1 states that training shared models on E2EE content is categorically incompatible with E2EE, but Recommendation 4 recommends opt-in consent as the "standard mechanism for allowing messaging services to train AI on user data" and explains how to design consent for such training. If the incompatibility is definitional, consent cannot change the security guarantee; if consent can make the processing acceptable in some circumstances, then the incompatibility in Recommendation 1 is not about E2EE confidentiality but about user agreement, and the paper should say so explicitly. The paper should clarify whether Recommendation 4 is a harm-reduction fallback for deployments that remain incompatible (which should be labeled as such) or a way to make training E2EE-compatible (which contradicts Recommendation 1).
minor comments (6)
  1. [§2.1.1, footnote 6] The informal gloss of semantic security as saying that "no computational adversary can guess any function of the message content" should be stated more precisely, e.g., "efficiently computable functions" and "with more than negligible advantage," to avoid overclaiming.
  2. [§4.2, FHE paragraph] The Zama estimate of roughly $5,000 per word is a vendor blog figure; it would be safer to cite peer-reviewed FHE performance measurements or to label the figure explicitly as a vendor-provided estimate, especially since the paper's argument does not depend on this particular number.
  3. [Table 1, row "On-device, fine-tuned model"] The table row marks the on-device configuration as fully compatible, but it omits the qualification stated in §4.2 Consideration #1 that endpoint-local training is compatible only if E2EE messages are used solely for the sender's and recipients' endpoint-local models and not for fine-tuning models on other devices.
  4. [§6.2.1] The discussion of FTC "precedent" should be framed as agency enforcement actions and consent decrees rather than judicial precedent, since the paper is describing FTC practice and investigation patterns rather than binding court decisions.
  5. [§7.3] The Apple Intelligence analysis depends on version-specific facts (e.g., iOS 18.1 vs. iOS 18.3 default settings); the paper should include a clear "status as of" date or a general reminder to re-check these fast-moving product details, since the cited states will age quickly.
  6. [§8.2, §4.2.1] There are several small presentation errors, including the duplicated "aligned aligned" in §8.2 and the stray spacing in the heading "F ederated learning" in §4.2.1; these should be corrected in a final pass.

Circularity Check

1 steps flagged · score 6.0 of 10

Recommendation 1 reduces by construction from the paper's stipulative 'derivative' definition of E2EE content, though independent memorization and extraction evidence and the rest of the framework keep the paper from being wholly circular.

  1. self definitional [Section 3.1 defines E2EE content; Section 4.2 applies the definition to shared models; Recommendation 1 in Section 7.1 draws the conclusion.]
    "E2EE content. We write E2EE content or E2EE data to refer to any data which a service provider explicitly or implicitly indicates is protected by E2EE in an application, and any derivatives of such data. ... So, training shared AI assistants with E2EE data definitionally undermines the confidentiality guarantees of E2EE technologies."

    Under this definition, a trained model is itself E2EE content, because Section 4.2 states 'a model is a derivative of its training data.' Shared-model outputs are then by definition 'derivatives of E2EE content,' and Section 2.1.1's confidentiality guarantee forbids third parties from learning 'any derivative of the message content.' Recommendation 1 is therefore not an empirical discovery but a direct consequence of the paper's stipulative definitions; the paper even labels it 'definitionally undermines.' The independent memorization and extraction evidence would support a narrower leakage-based incompatibility, but the categorical claim as stated would hold under the paper's definitions even for a model that never leaked verbatim data.

full rationale

The paper is a conceptual policy analysis rather than an empirical derivation, so the main circularity risk is definitional. Recommendation 1 ('Using end-to-end encrypted content to train shared AI models is not compatible with E2EE') is defended by saying that a model is a derivative of its training data and that shared-model outputs are therefore derivatives of other users' E2EE content. That conclusion is built into the paper's own definitions: Section 3.1 defines E2EE content to include any derivative of protected data, and Section 2.1.1 defines E2EE confidentiality as prohibiting third parties from learning any derivative of message content. Once those definitions are accepted, Recommendation 1 is a tautology, and the paper itself uses the word 'definitionally.' The independent evidence cited by the paper, namely memorization and training-data extraction results, gives non-circular support for a weaker, leakage-based incompatibility, but the categorical claim does not depend on that evidence. A related limitation is flagged by the paper itself in Section 3.2: it permits 'caveat features' (such as link previews) that involve third-party processing of 'strictly limited types or quantities of plaintext-dependent content,' while acknowledging that 'strictly limited' is ambiguous and that no precise threshold is articulated. A URL in a message is as much a derivative of E2EE content as many model outputs, so the derivative definition cannot simultaneously forbid all shared-model training and permit caveat features without an unstated quantitative boundary. This does not make the entire paper circular: the technical taxonomy, legal analysis, and disclosure and consent recommendations are independent contributions. The score reflects that one central claim reduces by construction, while the rest of the paper retains substantial independent content and the authors are transparent about their working definitions.

Assumptions & free parameters 0 free parameters · 4 assumptions · 0 invented entities

No free parameters or invented entities; the paper is a conceptual and legal analysis. The axioms list captures the definitional and domain assumptions that the recommendations depend on.

assumptions (4)
  • domain assumption E2EE confidentiality prohibits any third party from learning any function or derivative of message content, not just the exact plaintext.
    Invoked in Sections 2.1.1 and 3.1 to define E2EE content; this broad derivative definition is what makes model outputs count as E2EE content.
  • domain assumption A trained AI model is a derivative of its training data such that its outputs can reveal that data.
    Used in Sections 4.1.1 and 7.1 to argue shared models trained on E2EE content can leak it; supported by memorization and extraction attack literature.
  • domain assumption Trusted execution environments provide a security guarantee that is different in kind from cryptographic E2EE and cannot be substituted without changing the guarantee.
    Central to Recommendation 2 and the treatment of Apple's Private Cloud Compute in Sections 2.4.2 and 7.3.
  • domain assumption The legal interpretations of US FTC consumer protection law and EU GDPR/ePrivacy apply to E2EE messaging providers in the manner described.
    Underlies the legal analysis in Section 6; the paper notes legal uncertainty and pending processes.

how reviews work

0 comments
Cite this review

Pith. "Pith review of How To Think About End-To-End Encryption and AI: Training, Processing, Disclosure, and Consent." pith.science (2026). https://pith.science/paper/RZ6HZP7Q

@misc{pith2026241220231,
  author       = {Pith},
  title        = {Pith review of: How To Think About End-To-End Encryption and AI: Training, Processing, Disclosure, and Consent},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/RZ6HZP7Q}},
  note         = {Machine review of arXiv:2412.20231}
}
read the original abstract

End-to-end encryption (E2EE) has become the gold standard for securing communications, bringing strong confidentiality and privacy guarantees to billions of users worldwide. However, the current push towards widespread integration of artificial intelligence (AI) models, including in E2EE systems, raises some serious security concerns. This work performs a critical examination of the (in)compatibility of AI models and E2EE applications. We explore this on two fronts: (1) the integration of AI "assistants" within E2EE applications, and (2) the use of E2EE data for training AI models. We analyze the potential security implications of each, and identify conflicts with the security guarantees of E2EE. Then, we analyze legal implications of integrating AI models in E2EE applications, given how AI integration can undermine the confidentiality that E2EE promises. Finally, we offer a list of detailed recommendations based on our technical and legal analyses, including: technical design choices that must be prioritized to uphold E2EE security; how service providers must accurately represent E2EE security; and best practices for the default behavior of AI features and for requesting user consent. We hope this paper catalyzes an informed conversation on the tensions that arise between the brisk deployment of AI and the security offered by E2EE, and guides the responsible development of new AI features.

Figures

Figures reproduced from arXiv: 2412.20231 by the authors.

Figure 1
Figure 1. A sender S and receiver R communicate using an end-to-end encrypted application hosted by a company (middle). Solid lines represent plaintexts, and dashed lines represent ciphertexts. S and R can read their messages on their devices; however, while a message is “in transit” between their devices, it is encrypted so that it is not readable to the intermediary platform P handling it on its servers (or indeed to anyone… view at source ↗
Figure 2
Figure 2. Inference and training of AI assistants happen continuously as a feedback loop. As users continue [PITH_FULL_IMAGE:figures/full_fig_p009_2.png] view at source ↗
Figure 3
Figure 3. Generating an output from a user’s input, differentiating between what is on-device and on cloud [PITH_FULL_IMAGE:figures/full_fig_p010_3.png] view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Apple’s three-tiered approach Case I (Local model). Apple Intelligence’s first processing method consists of a series of small AI models stored entirely on a user’s device [15]. While some of these models are bespoke for certain tasks (e.g., code completion), their cor…
Figure 5
Figure 5. Figure 5: Evaluation framework for implementations of AI assistants [PITH_FULL_IMAGE:figures/full_fig_p021_5.png]
Figure 6
Figure 6. Figure 6: WhatsApp’s disclosure upon a user’s first use of Meta AI as of 15 November 2024 [PITH_FULL_IMAGE:figures/full_fig_p038_6.png]
Figure 8
Figure 8. Figure 8: Meta AI’s disclaimer when typing “@MetaAI” on What￾sApp for iOS for the first time, as of 03/15/25 [PITH_FULL_IMAGE:figures/full_fig_p054_8.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

180 extracted references · 67 canonical work pages

  1. [1]

    In re Blackbaud, Inc., Customer Data Breach Litigation, 567 F. Supp. 3d 667 (D.S.C. 2021)

  2. [2]

    Keys under doormats: mandating insecurity by requiring government access to all data and communications

    Harold Abelson et al. “Keys under doormats: mandating insecurity by requiring government access to all data and communications”. In: Journal of Cybersecurity (2015)

  3. [3]

    Privacy and rationality in individual decision making

    A. Acquisti and J. Grossklags. “Privacy and rationality in individual decision making”. In: IEEE Security & Privacy 3.1 (2005), pp. 26–33. doi: 10.1109/MSP.2005.22

  4. [4]

    What Can Behavioral Economics Teach Us About Privacy?

    Alessandro Acquisti and Jen Grossklags. “What Can Behavioral Economics Teach Us About Privacy?” In: Digital Privacy: Theory, Technologies and Practices (2007). permalink: https://perma.cc/ 56BA-YXXQ

  5. [5]

    https://www.ftc.gov/policy/advocacy- research/tech- at- ftc/2024/02/ai- other- companies-quietly-changing-your-terms-service-could-be-unfair-or-deceptive

    AI (and other) Companies: Quietly Changing Your Terms of Service Could Be Unfair or Decep- tive. https://www.ftc.gov/policy/advocacy- research/tech- at- ftc/2024/02/ai- other- companies-quietly-changing-your-terms-service-could-be-unfair-or-deceptive . perma- link: https://perma.cc/A2ZP-DWV7. 2024

  6. [6]

    Europe needs regulatory certainty on AI

    Meta et. al. “Europe needs regulatory certainty on AI”. In: Open Letter (2024)

  7. [7]

    AI Assistants: Breaking the Promise of End-to-End Encryption? NYU DeTaIL Lab Blog

    Sam de Alfaro et al. AI Assistants: Breaking the Promise of End-to-End Encryption? NYU DeTaIL Lab Blog. Mar. 2025. url: https://nyudetail.substack.com/e2ee-ai-broken-promise

  8. [8]

    Apple Intelligence now features Image Playground, Genmoji, Writing Tools enhancements, seamless support for ChatGPT, and visual intelligence

    Apple. Apple Intelligence now features Image Playground, Genmoji, Writing Tools enhancements, seamless support for ChatGPT, and visual intelligence . https://www.apple.com/newsroom/2024/ 12/apple- intelligence- now- features- image- playground- genmoji- and- more/. permalink: https://perma.cc/N6XN-T4FQ. 2024

Show all 180 references
  1. [9]

    Apple Platform Security

    Apple. Apple Platform Security . Tech. rep. permalink: https://perma.cc/LSK7-6ASF . 2024. url: https://help.apple.com/pdf/security/en_US/apple-platform-security-guide.pdf

  2. [10]

    Hardware Integrity

    Apple. Hardware Integrity . https : / / security . apple . com / documentation / private - cloud - compute/hardwareintegrity. permalink: https://perma.cc/8WZH-CZE2. 2024

  3. [11]

    How to get Apple Intelligence on iPhone

    Apple. How to get Apple Intelligence on iPhone . permalink: https://perma.cc/KBC3-AMFX. 2024. url: https://support.apple.com/en-us/121115

  4. [12]

    iCloud data security overview

    Apple. iCloud data security overview . https://support.apple.com/en- us/102651 . permalink: https://perma.cc/EP89-TPED. 2024

  5. [13]

    Integrating your app with Siri and Apple Intelligence

    Apple. Integrating your app with Siri and Apple Intelligence . https : / / developer . apple . com / documentation / appintents / integrating - your - app - with - siri - and - apple - intelligence. permalink: https://perma.cc/W7D2-ZPBM. 2024. 42

  6. [14]

    Introducing Apple Intelligence, the personal intelligence system that puts powerful generative models at the core of iPhone, iPad, and Mac

    Apple. Introducing Apple Intelligence, the personal intelligence system that puts powerful generative models at the core of iPhone, iPad, and Mac . https : / / www . apple . com / newsroom / 2024 / 06 / introducing- apple- intelligence- for- iphone- ipad- and- mac. permalink: ...

  7. [15]

    Introducing Apple’s On-Device and Server Foundation Models

    Apple. Introducing Apple’s On-Device and Server Foundation Models . https://machinelearning. apple.com/research/introducing-apple-foundation-models . permalink: https://perma.cc/ UB4P-BLRJ. 2024

  8. [16]

    iOS & iPadOS 18.3 RC Release Notes

    Apple. iOS & iPadOS 18.3 RC Release Notes . https://developer.apple.com/documentation/ ios- ipados- release- notes/ios- ipados- 18_3- release- notes. permalink: https://perma. cc/BK24-5QP8. 2024

  9. [17]

    Legal — Ask Siri, Dictation & Privacy

    Apple. Legal — Ask Siri, Dictation & Privacy . permalink: https://perma.cc/83H3-ZFA7 . 2024. url: https://www.apple.com/legal/privacy/data/en/ask-siri-dictation

  10. [18]

    Use Apple Intelligence with Siri on iPhone

    Apple. Use Apple Intelligence with Siri on iPhone . https://support.apple.com/en- gb/guide/ iphone/iph17bafe0f6/ios. permalink: https://perma.cc/R9XW-VYGL. 2024

  11. [19]

    Use ChatGPT with Apple Intelligence on iPhone

    Apple. Use ChatGPT with Apple Intelligence on iPhone . https://support.apple.com/en- mo/ guide/iphone/iph00fd3c8c2/ios. permalink: https://perma.cc/UC3Q-WRCX. 2024

  12. [20]

    Private Cloud Compute: A new frontier for AI privacy in the cloud

    Apple Security Engineering and Architecture (SEAR), User Privacy, Core Operating Systems (Core OS), Services Engineering (ASE), and Machine Learning and AI (AIML). Private Cloud Compute: A new frontier for AI privacy in the cloud . https://security.apple.com/blog/private-cloud...

  13. [21]

    Americans and Privacy: Concerned, Confused and Feeling Lack of Control Over Their Personal Information

    Brooke Auxier et al. Americans and Privacy: Concerned, Confused and Feeling Lack of Control Over Their Personal Information . Pew Research Center, 2019

  14. [22]

    Processing of Personal Data for AI Training in Brazil: Takeaways from ANPD’s Preliminary Decisions in the Meta Case

    Maria Badillo. “Processing of Personal Data for AI Training in Brazil: Takeaways from ANPD’s Preliminary Decisions in the Meta Case”. In: Future of Privacy Forum (2024). permalink: https: //perma.cc/24GE-ABNN

  15. [23]

    Rethinking Encryption

    Jim Baker. Rethinking Encryption . https : / / www . lawfaremedia . org / article / rethinking - encryption. permalink: https://perma.cc/H2XZ-3SU7. 2019

  16. [24]

    Information Fiduciaries and the First Amendment

    Jack M. Balkin. “Information Fiduciaries and the First Amendment”. In: UC Davis Law Review (2016). permalink: https://perma.cc/7MD4-HV79

  17. [25]

    Cryptography, Trust and Privacy: It’s Complicated

    Ero Balsa, Helen Nissenbaum, and Sunoo Park. “Cryptography, Trust and Privacy: It’s Complicated”. In: CSLA W. Ed. by Daniel J. Weitzner, Joan Feigenbaum, and Christopher S. Yoo. 2022

  18. [26]

    Searching for the Common Law: The Quantitative Approach of the Restatement of Consumer Contracts

    Oren Bar-Gill, Omri Ben-Shahar, and Florencia Marotta-Wurgler. “Searching for the Common Law: The Quantitative Approach of the Restatement of Consumer Contracts”. In: University of Chicago Law Review: Vol. 84: Iss. 1, Article 2 (2017). permalink: https://perma.cc/5MF2-E7WP

  19. [27]

    Going Bright: Wiretapping without Weakening Communications Infras- tructure

    Steven M. Bellovin et al. “Going Bright: Wiretapping without Weakening Communications Infras- tructure”. In: IEEE Security & Privacy (2013)

  20. [28]

    It’s Too Complicated: How the Internet Upends Katz, Smith, and Electronic Surveillance Law

    Steven M Bellovin et al. “It’s Too Complicated: How the Internet Upends Katz, Smith, and Electronic Surveillance Law”. In: Harv. JL & Tech. (2016)

  21. [29]

    Apple Sees Value in Its Stand to Protect Security

    Katie Benner and Paul Mozur. Apple Sees Value in Its Stand to Protect Security . https://www. nytimes.com/2016/02/21/technology/apple-sees-value-in-privacy-vow.html . permalink: https://perma.cc/8QR5-KWLD. 2016

  22. [30]

    How Tim Cook, in iPhone Battle, Became a Bulwark for Digital Privacy

    Katie Benner and Nicole Perlroth. How Tim Cook, in iPhone Battle, Became a Bulwark for Digital Privacy. https : / / www . nytimes . com / 2016 / 02 / 19 / technology / how - tim - cook - became - a - bulwark-for-digital-privacy.html . permalink: https://perma.cc/F3AF-N6UM. 2016

  23. [31]

    Guidelines 03/2022 on Deceptive design patterns in social media platform interfaces: how to recognise and avoid them

    European Data Protection Board. “Guidelines 03/2022 on Deceptive design patterns in social media platform interfaces: how to recognise and avoid them”. In: (2022). permalink: https://perma.cc/ RLY5-FN2X. 43

  24. [32]

    Opinion 28/2024 on certain data protection aspects related to the processing of personal data in the context of AI models

    European Data Protection Board. “Opinion 28/2024 on certain data protection aspects related to the processing of personal data in the context of AI models”. In: (2024). permalink: https://perma. cc/MB7M-KMHF

  25. [33]

    Report of the work undertaken by the ChatGPT Taskforce

    European Data Protection Board. “Report of the work undertaken by the ChatGPT Taskforce”. In: (2024). permalink: https://perma.cc/J6UQ-ZM82

  26. [34]

    Statement 03/2021 on the ePrivacy Regulation

    European Data Protection Board. “Statement 03/2021 on the ePrivacy Regulation”. In: (2021). permalink: https://perma.cc/92MS-2HXJ

  27. [35]

    Reinvigorating Consent: Exploring New Paradigms for Privacy and Data Sharing

    August Bourgeus and Laurens Vandercruysse. “Reinvigorating Consent: Exploring New Paradigms for Privacy and Data Sharing”. In: Proceedings of 22nd European Conference on Computer-Supported Cooperative Work . European Society for Socially Embedded Technologies (EUSSET), 2024. d...

  28. [36]

    The Shape of Consumer Contracts

    Andrea J. Boyack. “The Shape of Consumer Contracts”. In: Denver Law Review (2023). permalink: https://perma.cc/T4LR-6CR5

  29. [37]

    The Brussels Effect: How the European Union Rules the World

    Anu Bradford. “The Brussels Effect: How the European Union Rules the World”. In: Oxford University Press (2020). permalink: https://perma.cc/4AND-6QJB

  30. [38]

    Software Grand Exposure: SGX Cache Attacks Are Practical

    Ferdinand Brasser et al. “Software Grand Exposure: SGX Cache Attacks Are Practical”. In: WOOT. 2017

  31. [39]

    Deceptive patterns – user interfaces designed to trick you

    H Brignull et al. Deceptive patterns – user interfaces designed to trick you . https://www.deceptive. design/. permalink: https://perma.cc/KQ4T-F7QD. Apr. 2023

  32. [40]

    What does it mean for a language model to preserve privacy?

    Hannah Brown et al. “What does it mean for a language model to preserve privacy?” In: F AccT. 2022

  33. [41]

    Deniable encryption

    Rein Canetti et al. “Deniable encryption”. In: CRYPTO. 1997

  34. [42]

    Extracting Training Data from Large Language Models

    Nicholas Carlini et al. “Extracting Training Data from Large Language Models”. In: 30th USENIX Security Symposium (USENIX Security 21) . 2021

  35. [43]

    EU ditches plans to regulate tech patents, AI liability, online privacy

    Foo Yun Chee. “EU ditches plans to regulate tech patents, AI liability, online privacy”. In: Reuters (2025). url: https : / / www . reuters . com / technology / eu - ditches - plans - regulate - tech - patents-ai-liability-online-privacy-2025-02-12/

  36. [44]

    Adverse impacts of revealing the presence of “Artificial Intelligence (AI)

    Mesut Cicek, Dogan Gursoy, and Lu Lu. “Adverse impacts of revealing the presence of “Artificial Intelligence (AI)” technology in product and service descriptions on purchase intentions: the mediating role of emotional trust and the moderating role of perceived risk”. In:Journa...

  37. [45]

    Europe’s privacy patrol is spoiling Big Tech’s AI party

    Sam Clark and Pieter Haeck. Europe’s privacy patrol is spoiling Big Tech’s AI party . https : / / www . politico . eu / article / europe - privacy - patrol - vengeance - block - ai - artificial - intelligence/. permalink: https://perma.cc/A8YF-MM3U. 2024

  38. [46]

    Mortgage Company Settles Information Security Charges

    Jessica Rich Claudia Bourne Farrell. Mortgage Company Settles Information Security Charges. https: //www.ftc.gov/news- events/news/press- releases/2005/09/mortgage- company- settles- information-security-charges. permalink: https://perma.cc/GM74-8VZS. 2005

  39. [47]

    The Privacy Fallacy

    Ignacio Cofone. “The Privacy Fallacy”. In: Cambridge University Press (2023). permalink: https: //perma .cc/L4D8- TSFV. url: https:/ /www.cambridge.org /core/books/ privacy- fallacy/ 547578F2A1AE0C40963105CE066B412E

  40. [48]

    Examined Lives: Informational Privacy and the Subject as Object

    Julie E. Cohen. “Examined Lives: Informational Privacy and the Subject as Object”. In: 52 Stan. L. Rev. 1373-1438 (2000). permalink: https://perma.cc/VR6L-87AS

  41. [49]

    Julie E. Cohen. Law for the Platform Economy . https : / / scholarship . law . georgetown . edu / facpub/2015. permalink: https://perma.cc/59AP-9F32. 2017

  42. [50]

    What Privacy Is For

    Julie E. Cohen. “What Privacy Is For”. In: Georgetown Law Faculty Publications and Other Works (2013). permalink: https://perma.cc/WQU4-PAJV

  43. [51]

    On post-compromise security

    Katriel Cohn-Gordon, Cas Cremers, and Luke Garratt. “On post-compromise security”. In: IEEE CSF. 2016. 44

  44. [52]

    Data Protection Commission welcomes conclusion of proceedings relat- ing to X’s AI tool ‘Grok’

    Data Protection Commission. Data Protection Commission welcomes conclusion of proceedings relat- ing to X’s AI tool ‘Grok’ . https://www.dataprotection.ie/en/news- media/press- releases/ data - protection - commission - welcomes - conclusion - proceedings - relating - xs - ai ...

  45. [53]

    The DPC’s Engagement with Meta on AI

    Data Protection Commission. The DPC’s Engagement with Meta on AI. https://www.dataprotection. ie/en/news-media/latest-news/dpcs-engagement-meta-ai . permalink: https://perma.cc/ Y65M-U72C. 2024

  46. [54]

    Consumer protection: manipulative online practices found on 148 out of 399 online shops screened

    European Commission. Consumer protection: manipulative online practices found on 148 out of 399 online shops screened . https://ec.europa.eu/commission/presscorner/detail/en/ip_23_418 . permalink: https://perma.cc/J24T-LCEH. 2023

  47. [55]

    Mergers: Commission approves acquisition of WhatsApp by Facebook

    European Commission. “Mergers: Commission approves acquisition of WhatsApp by Facebook”. In: Press Release (2014). permalink: https://perma.cc/5EU4-KNWE

  48. [56]

    Mergers: Commission fines Facebook €110 million for providing misleading information about WhatsApp takeover

    European Commission. “Mergers: Commission fines Facebook €110 million for providing misleading information about WhatsApp takeover”. In: Press Release (2016). permalink: https://perma.cc/ V9MC-YHGQ

  49. [57]

    $20m penalty for Meta companies for conduct liable to mislead consumers about use of their data

    Australia Competition and Consumer Commission. $20m penalty for Meta companies for conduct liable to mislead consumers about use of their data . https://www.accc.gov.au/media- release/ 20m-penalty-for-meta-companies-for-conduct-liable-to-mislead-consumers-about-use- of-their-d...

  50. [58]

    The Files are in the Computer: Copyright, Memorization, and Generative AI

    A Feder Cooper and James Grimmelmann. “The Files are in the Computer: Copyright, Memorization, and Generative AI”. In: arXiv preprint arXiv:2404.12590 (2024)

  51. [59]

    Intel SGX Explained

    Victor Costan and Srinivas Devadas. “Intel SGX Explained”. In: Cryptology ePrint Archive (2016)

  52. [60]

    Launching Default End-to-End Encryption on Messenger

    Loredana Crisan. Launching Default End-to-End Encryption on Messenger . https : / / about . fb . com / news / 2023 / 12 / default - end - to - end - encryption - on - messenger. permalink: https : //perma.cc/9HSW-WNFG. 2023

  53. [61]

    Enhanced Visibility and Harden- ing Guidance for Communications Infrastructure

    Cybersecurity and Infrastructure Security Agency (CISA) et al. Enhanced Visibility and Harden- ing Guidance for Communications Infrastructure . https : / / www . cisa . gov / resources - tools / resources/enhanced-visibility-and-hardening-guidance-communications-infrastructure...

  54. [62]

    https://www.gov.br/anpd/pt- br/assuntos/ noticias/anpd- determina- suspensao- cautelar- do- tratamento- de- dados- pessoais- para- treinamento-da-ia-da-meta

    Autoridade Nacional de Prote¸ c˜ ao de Dados.ANPD determina suspens˜ ao cautelar do tratamento de dados pessoais para treinamento da IA da Meta . https://www.gov.br/anpd/pt- br/assuntos/ noticias/anpd- determina- suspensao- cautelar- do- tratamento- de- dados- pessoais- para- ...

  55. [63]

    https://www.gov

    Autoridade Nacional de Prote¸ c˜ ao de Dados.Meta cumpre exigˆ encias da ANPD e poder´ a retomar, com restri¸ c˜ oes, o uso de dados pessoais para treinamento de inteligˆ encia artificial. https://www.gov. br/anpd/pt- br/assuntos/noticias/meta- cumpre- exigencias- da- anpd- e-...

  56. [64]

    Understanding Apple’s On-Device and Server Foundation Models release

    Artem Dinaburg. Understanding Apple’s On-Device and Server Foundation Models release . https: / / blog . trailofbits . com / 2024 / 06 / 14 / understanding - apples - on - device - and - server - foundations-model-release/. permalink: https://perma.cc/AFW8-LXND. 2024

  57. [65]

    Anonymity Loves Company: Usability and the Network Ef- fect

    Roger Dingledine and Nick Mathewson. “Anonymity Loves Company: Usability and the Network Ef- fect”. In: 5th Annual Workshop on the Economics of Information Security, WEIS 2006, Robinson Col- lege, University of Cambridge, England, UK, June 26-28, 2006 . 2006. url: https://econ...

  58. [66]

    Discord’s audio/video end-to-end encryption (DA VE) protocol

    Discord. “Discord’s audio/video end-to-end encryption (DA VE) protocol”. In: (2024)

  59. [67]

    How Companies Learn Your Secrets

    Charles Duhigg. How Companies Learn Your Secrets . https://www.nytimes.com/2012/02/19/ magazine/shopping-habits.html. 2012. 45

  60. [68]

    Differential Privacy: A Survey of Results

    Cynthia Dwork. “Differential Privacy: A Survey of Results”. In: Theory and Applications of Models of Computation, 5th International Conference, TAMC 2008, Xi’an, China, April 25-29, 2008. Pro- ceedings. Ed. by Manindra Agrawal et al. Vol. 4978. Lecture Notes in Computer Scienc...

  61. [69]

    So far, AI hasn ’t been profitable for Big Tech

    Benj Edwards. So far, AI hasn ’t been profitable for Big Tech. https://arstechnica.com/information- technology/2023/10/so-far-ai-hasnt-been-profitable-for-big-tech . permalink: https: //perma.cc/B3R3-GA2Y

  62. [70]

    South Korea fines Meta €14 million for illegally collecting information on Facebook users

    Euronews. “South Korea fines Meta €14 million for illegally collecting information on Facebook users”. In: Euronews Tech News (2024). permalink: https://perma.cc/8MY9-X4MP

  63. [71]

    Security Vulnerabilities of SGX and Countermeasures: A Survey

    Shufan Fei et al. “Security Vulnerabilities of SGX and Countermeasures: A Survey”. In: ACM Com- puting Surveys (CSUR) (2021)

  64. [72]

    Geoffrey A. Fowler. I tried to read all my app privacy policies. It was 1 million words. https : //www.washingtonpost.com/technology/2022/05/31/abolish-privacy-policies/. 2022

  65. [73]

    Building AI Technology for Europeans in a Transparent and Responsible Way

    Stefano Fratta. Building AI Technology for Europeans in a Transparent and Responsible Way. https: //about.fb.com/news/2024/06/building-ai-technology-for-europeans-in-a-transparent- and-responsible-way/. permalink: https://perma.cc/6377-A64J. 2024

  66. [74]

    Trump picks Andrew Ferguson to chair FTC

    Jody Godoy. “Trump picks Andrew Ferguson to chair FTC”. In: Reuters (2024)

  67. [75]

    Corrective machine unlearning

    Shashwat Goel et al. “Corrective machine unlearning”. In: arXiv preprint arXiv:2402.14015 (2024)

  68. [76]

    Use end-to-end encryption in Google Messages

    Google. Use end-to-end encryption in Google Messages . permalink: https://perma.cc/WP25-WD89

  69. [77]

    The Vital Role of End-to-End Encryption

    Jennifer Stisa Granick and Daniel Kahn Gillmor. The Vital Role of End-to-End Encryption . https: //www.aclu.org/news/privacy- technology/the- vital- role- of- end- to- end- encryption. permalink: https://perma.cc/RN4P-K94X

  70. [78]

    Google’s A.I

    Nico Grant. Google’s A.I. Search Errors Cause a Furor Online. https://www.nytimes.com/2024/05/ 24/technology/google-ai-overview-search.html . permalink: https://perma.cc/NG3P-JFM5

  71. [79]

    Let’s talk about AI and end-to-end encryption

    Matthew Green. Let’s talk about AI and end-to-end encryption . A Few Thoughts on Cryptographic Engineering. Jan. 2025. url: https://blog.cryptographyengineering.com/2025/01/17/lets- talk-about-ai-and-end-to-end-encryption

  72. [80]

    Looking back at the Snowden revelations

    Matthew Green. Looking back at the Snowden revelations. https://blog.cryptographyengineering. com/2019/09/24/looking-back-at-the-snowden-revelations . permalink: https://perma.cc/ F6PV-SW9H

  73. [81]

    Forward secure asynchronous messaging from puncturable encryp- tion

    Matthew D Green and Ian Miers. “Forward secure asynchronous messaging from puncturable encryp- tion”. In: IEEE S&P . 2015

  74. [82]

    Hacker Lexicon: What Is End-to-End Encryption? https://www.wired.com/2014/ 11/hacker-lexicon-end-to-end-encryption/

    Andy Greenberg. Hacker Lexicon: What Is End-to-End Encryption? https://www.wired.com/2014/ 11/hacker-lexicon-end-to-end-encryption/ . permalink: https://perma.cc/JET4-3P8T. 2014

  75. [83]

    Signal Is More Than Encrypted Messaging

    Andy Greenberg. Signal Is More Than Encrypted Messaging. Under Meredith Whittaker, It’s Out to Prove Surveillance Capitalism Wrong . https://www.wired.com/story/meredith- whittaker- signal/. permalink: https://perma.cc/6JMM-537E. 2024

  76. [84]

    On end-to-end encryption

    Britta Hale and Chelsea Komlo. “On end-to-end encryption”. In: Cryptology ePrint Archive (2022)

  77. [85]

    Top EU privacy regulator opens probe into Google’s AI compliance

    Padraic Halpin. Top EU privacy regulator opens probe into Google’s AI compliance. http://archive. today/2024.09.12-004248/https://www.reuters.com/technology/artificial-intelligence/ top - eu - privacy - regulator - opens - probe - into - googles - ai - compliance - 2024 - 09 -...

  78. [86]

    Nobody knows how AI works

    Melissa Heikkil¨ a. Nobody knows how AI works . https://www.technologyreview.com/2024/03/05/ 1089449/nobody-knows-how-ai-works/ . permalink: https://perma.cc/XTN9-J5DA. 2024

  79. [87]

    How Target Figured Out A Teen Girl Was Pregnant Before Her Father Did

    Kashmir Hill. How Target Figured Out A Teen Girl Was Pregnant Before Her Father Did . https: //www.forbes.com/sites/kashmirhill/2012/02/16/how-target-figured-out-a-teen-girl- was-pregnant-before-her-father-did/ . permalink: https://perma.cc/D944-EEFW. 2022. 46

  80. [88]

    Homomorphic Encryption and LLM : Is ChatGPT end to end encrypted ? permalink: https://perma.cc/4TQZ-7ADT

    Rand Hindi. Homomorphic Encryption and LLM : Is ChatGPT end to end encrypted ? permalink: https://perma.cc/4TQZ-7ADT . 2023. url: https://www.zama.ai/post/chatgpt-privacy-with- homomorphic-encryption

  81. [89]

    Alan Westin’s Privacy Homo Economicus

    Chris Jay Hoofnagle and Jennifer M. Urban. “Alan Westin’s Privacy Homo Economicus”. In: 49 Wake Forest Law Review 261 (2014), UC Berkeley Public Law Research Paper No. 2434800 (2014). permalink: https://perma.cc/2EWE-7DU9

  82. [90]

    https://faq.whatsapp.com/481135090640375/ ?cms_platform=android

    How to back up your chat history | WhatsApp Help Center. https://faq.whatsapp.com/481135090640375/ ?cms_platform=android. permalink: https://perma.cc/AP3F-9X6S. (Visited on 01/15/2025)

  83. [91]

    Membership inference attacks on machine learning: A survey

    Hongsheng Hu et al. “Membership inference attacks on machine learning: A survey”. In: ACM Com- puting Surveys (CSUR) 54.11s (2022), pp. 1–37

  84. [92]

    A beginner’s guide to EU rules on scanning private communications: Part 1

    Ella Jakubowska. A beginner’s guide to EU rules on scanning private communications: Part 1 . https : / / edri . org / our - work / a - beginners - guide - to - eu - rules - on - scanning - private - communications-part-1/. permalink: https://perma.cc/63ZP-2D44. 2021

  85. [93]

    Dental Practice Software Provider Settles FTC Charges It Misled Customers About Encryption of Patient Data

    Kristin Madigan Jay Mayfield Jessica Lyon. Dental Practice Software Provider Settles FTC Charges It Misled Customers About Encryption of Patient Data . https://www.ftc.gov/news-events/news/ press- releases/2016/01/dental- practice- software- provider- settles- ftc- charges- it...

  86. [94]

    AMD memory encryption

    David Kaplan, Jeremy Powell, and Tom Woller. AMD memory encryption . https://www.amd.com/ content / dam / amd / en / documents / epyc - business - docs / white - papers / memory - encryption - white-paper.pdf. permalink: https://perma.cc/WY82-655P. 2021

  87. [95]

    Report on encryption, anonymity, and the human rights framework

    David Kaye. Report on encryption, anonymity, and the human rights framework . https : / / www . undocs.org/A/HRC/29/32. permalink: https://perma.cc/QN7A-R9V5. 2015

  88. [96]

    FTC Takes Action Against Marriott and Starwood Over Multiple Data Breaches

    Rebecca Kern. “FTC Takes Action Against Marriott and Starwood Over Multiple Data Breaches”. In: (2024)

  89. [97]

    Growing public concern about the role of artificial intelligence in daily life

    Alec Tyson and Emma Kikuchi. Growing public concern about the role of artificial intelligence in daily life . https://www.pewresearch.org/short-reads/2023/08/28/growing-public-concern- about-the-role-of-artificial-intelligence-in-daily-life/ . permalink: https://perma. cc/3BDG...

  90. [98]

    The FBI Wants You to Know it Has Not Changed its Position on Encryption

    Mallory Knodel. The FBI Wants You to Know it Has Not Changed its Position on Encryption. And That’s a Problem. Tech Policy Press. permalink: https://perma.cc/7ZHL-54WU. Dec. 2024

  91. [99]

    Tech Policy Press

    Mallory Knodel and Andr´ es F´ abrega.Can Bots Read Your Encrypted Messages? Encryption, Privacy, and the Emerging AI Dilemma . Tech Policy Press. Feb. 2025. url: https : / / www . techpolicy . press/can- bots- read- your- encrypted- messages- encryption- privacy- and- the- em...

  92. [100]

    Definition of End-to-end Encryption

    Mallory Knodel et al. Definition of End-to-end Encryption . https://datatracker.ietf.org/doc/ html/draft-knodel-e2ee-definition-11 . permalink: https://perma.cc/33J2-TKNE. 2023

  93. [101]

    Inferring Fine-grained Control Flow Inside SGX Enclaves with Branch Shadowing

    Sangho Lee et al. “Inferring Fine-grained Control Flow Inside SGX Enclaves with Branch Shadowing”. In: USENIX Security. 2017

  94. [102]

    Privacy and Information Sharing

    Maeve Duggan Lee Rainie. Privacy and Information Sharing. 2016. url: https://www.pewresearch. org/internet/2016/01/14/privacy-and-information-sharing/

  95. [103]

    Code and Other Laws of Cyberspace

    Lawrence Lessig. Code and Other Laws of Cyberspace . Basic Books, 1999. isbn: 0-465-03912-X

  96. [104]

    How Generative AI Is Making Customer Experience Worse

    Tom Lewis. “How Generative AI Is Making Customer Experience Worse”. In: Forbes (2024). perma- link: https://perma.cc/44FY-QYFV

  97. [105]

    FTC Requires Zoom to Enhance its Security Practices as Part of Settlement

    Caroline Schmitz Linda Holleran Kopp Ryan Mehm. “FTC Requires Zoom to Enhance its Security Practices as Part of Settlement”. In: (2020). permalink: https://perma.cc/24RA-7NB5

  98. [106]

    Group Privacy, New Challenges of Data Technolo- gies

    Bart van der Sloot Linnet Taylor Luciano Floridi. Group Privacy, New Challenges of Data Technolo- gies. Springer Cham, 2017. isbn: 978-3-319-46606-4. 47

  99. [107]

    We Read 150 Privacy Policies

    Kevin Litman-Navarro. We Read 150 Privacy Policies. They Were an Incomprehensible Disaster. https://www.nytimes.com/interactive/2019/06/12/opinion/facebook- google- privacy- policies.html. permalink: https://perma.cc/5XJJ-4CQV. 2019

  100. [108]

    I link therefore I am

    Joshua Lund. I link therefore I am . https : / / signal . org / blog / i - link - therefore - i - am/. permalink: https://perma.cc/9DZY-2BHG. 2019

  101. [109]

    Meta to European Union: Your Tech Rules Threaten to Squelch the AI Boom

    Kim Mackrael. “Meta to European Union: Your Tech Rules Threaten to Squelch the AI Boom”. In: The Washington Post (2024)

  102. [110]

    Technology preview: Private contact discovery for Signal

    Moxie Marlinspike. Technology preview: Private contact discovery for Signal . https://signal.org/ blog/private-contact-discovery/. permalink: https://perma.cc/DR7X-3G2E. 2017

  103. [111]

    Will Increased Disclosure Help? Evaluating the Recommendations of the ALI’s Principles of the Law of Software Contracts

    Florencia Marotta-Wurgler. “Will Increased Disclosure Help? Evaluating the Recommendations of the ALI’s Principles of the Law of Software Contracts”. In: University of Chicago Law Review: Vol. 78: Iss. 1, Article 9 (2011). permalink: https://perma.cc/BP4K-VPTQ

  104. [112]

    What is retrieval-augmented generation? https : / / research

    Kim Martineau. What is retrieval-augmented generation? https : / / research . ibm . com / blog / retrieval-augmented-generation-RAG . permalink: https://perma.cc/48JR-WJ2U . Aug. 2023. (Visited on 12/11/2024)

  105. [113]

    (Anti)Trust Issues

    Elaine McArdle. “(Anti)Trust Issues”. In: Harvard Law Bulletin (2024). permalink: https://perma. cc/HVX2-742Z

  106. [114]

    How much do language models copy from their training data? evaluating lin- guistic novelty in text generation using raven

    R Thomas McCoy et al. “How much do language models copy from their training data? evaluating lin- guistic novelty in text generation using raven”. In: Transactions of the Association for Computational Linguistics 11 (2023), pp. 652–670

  107. [115]

    The cost of reading privacy policies

    Aleecia M McDonald and Lorrie Faith Cranor. “The cost of reading privacy policies”. In: I/S: A Journal of Law & Policy for the Information Society 4 (2009), pp. 543–897

  108. [116]

    Building AI Technology for the UK in a Responsible and Transparent Way

    Meta. “Building AI Technology for the UK in a Responsible and Transparent Way”. In: Newsroom for Europe, Middle East and Africa (2024). permalink: https://perma.cc/VGZ6-QGHK

  109. [117]

    How Meta uses information for generative AI models and features

    Meta. How Meta uses information for generative AI models and features . permalink: https : / / perma.cc/57AG-YUNX. 2025. url: https://www.facebook.com/privacy/genai

  110. [118]

    The future of AI: Built with Llama

    Meta. The future of AI: Built with Llama . permalink: https://perma.cc/2GRP-7MD4 . 2024. url: https://ai.meta.com/blog/future-of-ai-built-with-llama/

  111. [119]

    Miller III

    James C. Miller III. FTC Policy Statement on Deception . https://www.ftc.gov/system/files/ documents/public_statements/410531/831014deceptionstmt.pdf . Letter. permalink: https: //perma.cc/XD4D-AML3

  112. [120]

    Elon Musk’s X under pressure from regulators over data harvesting for Grok AI

    Dan Milmo. Elon Musk’s X under pressure from regulators over data harvesting for Grok AI . https: //www.theguardian.com/technology/article/2024/jul/26/elon-musks-x-under-pressure- from-regulators-over-data-harvesting-for-grok-ai . permalink: https://perma.cc/SPP5- QD5S. 2024

  113. [121]

    Anti-trust regulator levy Rs 213 crore penalty on Meta for ’forcing’ WhatsApp users to share data

    Dipak Mondal. “Anti-trust regulator levy Rs 213 crore penalty on Meta for ’forcing’ WhatsApp users to share data”. In: The New Indian Express (2024). permalink: https://perma.cc/863A-8AVC

  114. [122]

    Scalable extraction of training data from (production) language models

    Milad Nasr et al. “Scalable extraction of training data from (production) language models”. In: arXiv preprint arXiv:2311.17035 (2023)

  115. [123]

    Decrypting the Encryption Debate: A Framework for Decision Makers

    National Academies of Sciences, Engineering, and Medicine. Decrypting the Encryption Debate: A Framework for Decision Makers . permalink: https : / / perma . cc / NX82 - EMD8. The National Academies Press, 2018. url: https://nap.nationalacademies.org/catalog/25010/decrypting- ...

  116. [124]

    Privacy in Context: Technology, Policy and the Integrity of Social Life

    Helen Nissenbaum. Privacy in Context: Technology, Policy and the Integrity of Social Life . Stanford Law Books, 2010. isbn: 0804752370

  117. [125]

    Dark Patterns after the GDPR: Scraping Consent Pop-ups and Demonstrat- ing their Influence

    Midas Nouwens et al. “Dark Patterns after the GDPR: Scraping Consent Pop-ups and Demonstrat- ing their Influence”. In: Proceedings of the 2020 CHI Conference on Human Factors in Computing Systems. CHI ’20. ACM, Apr. 2020. doi: 10.1145/3313831.3376321 . url: http://dx.doi.org/ ...

  118. [126]

    noyb urges 11 DPAs to immediately stop Meta’s abuse of personal data for AI

    NOYB. “noyb urges 11 DPAs to immediately stop Meta’s abuse of personal data for AI”. In: Forced Consent & Consent Bypass (2024). permalink: https://perma.cc/92R9-XX5D

  119. [127]

    Smart Privacy for Web3 & AI

    Oasis Labs. Smart Privacy for Web3 & AI . permalink: https://perma.cc/9CXH-8964. url: \url{https://oasisprotocol.org/}

  120. [128]

    Due diligence when sharing data following mergers and acquisi- tions

    Information Commissioner’s Office. “Due diligence when sharing data following mergers and acquisi- tions”. In: For Organizations (). permalink: https://perma.cc/WMJ8-W8AJ

  121. [129]

    ICO statement in response to Meta’s announcement on user data to train AI

    Information Commissioner’s Office. ICO statement in response to Meta’s announcement on user data to train AI . https://ico.org.uk/about-the-ico/media-centre/news-and-blogs/2024/09/ico- statement- in- response- to- metas- announcement- on- user- data- to- train- ai/. permalink:...

  122. [130]

    When can we rely on legitimate interests? https://ico.org

    Information Commissioner’s Office. When can we rely on legitimate interests? https://ico.org. uk / for - organisations / uk - gdpr - guidance - and - resources / lawful - basis / legitimate - interests / when - can - we - rely - on - legitimate - interests/. permalink: https :...

  123. [131]

    Data Controls F AQ

    OpenAI. Data Controls F AQ. https://help.openai.com/en/articles/7730893-data-controls- faq. WebArchive: https://web.archive.org/web/20250109132809/https://help.openai. com/en/articles/7730893-data-controls-faq . 2024

  124. [132]

    How your data is handled when you use ChatGPT through Apple’s integrations

    OpenAI. How your data is handled when you use ChatGPT through Apple’s integrations . https : / / help . openai . com / en / articles / 9737562 - how - your - data - is - handled - when - you - use - chatgpt - through - apple - s - integrations. WebArchive: https : / / web . ar...

  125. [133]

    Setting up ChatGPT with Apple Intelligence

    OpenAI. Setting up ChatGPT with Apple Intelligence . https://help.openai.com/en/articles/ 10269382- setting- up- chatgpt- with- apple- intelligence. permalink: https://perma.cc/ V8M6-68XK. 2024

  126. [134]

    Training language models to follow instructions with human feedback

    Long Ouyang et al. “Training language models to follow instructions with human feedback”. In: Advances in neural information processing systems 35 (2022), pp. 27730–27744

  127. [135]

    Understanding the privacy-personalization dilemma for web search: a user perspective

    Saurabh Panjwani et al. “Understanding the privacy-personalization dilemma for web search: a user perspective”. In: 2013 ACM SIGCHI Conference on Human Factors in Computing Systems, CHI ’13, Paris, France, April 27 - May 2, 2013 . Ed. by Wendy E. Mackay, Stephen A. Brewster, a...

  128. [136]

    How to deploy machine learning with differen- tial privacy

    Nicolas Papernot and Abhradeep Guha Thakurta. How to deploy machine learning with differen- tial privacy . https://www.nist.gov/blogs/cybersecurity- insights/how- deploy- machine- learning-differential-privacy. permalink: https://perma.cc/39ZU-CGBV. 2021

  129. [137]

    Proposal for a regulation on privacy and electronic communications

    European Parliament. Proposal for a regulation on privacy and electronic communications . https: //www.europarl.europa.eu/legislative- train/theme- connected- digital- single- market/ file-jd-e-privacy-reform . permalink: https://perma.cc/796G-R3J7. 2024

  130. [138]

    Formal abstractions for attested execution secure processors

    Rafael Pass, Elaine Shi, and Florian Tramer. “Formal abstractions for attested execution secure processors”. In: Eurocrypt. 2017

  131. [139]

    The Double Rachet Algorithm

    Trevor Perin and Moxie Marlinspike. The Double Rachet Algorithm . permalink: https://perma. cc / H26Z - TYD2. 2016. url: https : / / signal . org / docs / specifications / doubleratchet / doubleratchet.pdf

  132. [141]

    How to DP-fy ML: A Practical Guide to Machine Learning with Dif- ferential Privacy

    Natalia Ponomareva et al. “How to DP-fy ML: A Practical Guide to Machine Learning with Dif- ferential Privacy”. In: Journal of Artificial Intelligence Research 77 (July 2023), 1113–1201. issn: 1076-9757. doi: 10.1613/jair.1.14649. url: http://dx.doi.org/10.1613/jair.1.14649. 49

  133. [142]

    Apple ordered to open encrypted user accounts globally to UK spying

    Dominic Preston. Apple ordered to open encrypted user accounts globally to UK spying . permalink: https://perma.cc/NM5B-77PN . 2025. url: https://www.theverge.com/news/608145/apple-uk- icloud-encrypted-backups-spying-snoopers-charter

  134. [143]

    India restricts WhatsApp sharing data with other Meta entities, imposes $25.4 mln fine

    Reuters. “India restricts WhatsApp sharing data with other Meta entities, imposes $25.4 mln fine”. In: Reuters - Boards, Policy & Regulation (2024)

  135. [144]

    Change the default messaging app on your Galaxy phone

    Samsung. Change the default messaging app on your Galaxy phone . permalink: https://perma. cc/XZV9-N9QR. 2025. url: https://www.samsung.com/us/support/answer/ANS10001947/

  136. [145]

    Official Replay — Galaxy Unpacked January 2025 — Samsung

    Samsung. Official Replay — Galaxy Unpacked January 2025 — Samsung . permalink: https:// perma.cc/JDP6-35KU . 2025. url: https://www.youtube.com/watch?v=HinL5jCy_oI&t=608s&ab_ channel=Samsung

  137. [146]

    Samsung and Google Cloud Join Forces to Bring Generative AI to Samsung Galaxy S24 series

    Samsung. Samsung and Google Cloud Join Forces to Bring Generative AI to Samsung Galaxy S24 series. permalink: https://perma.cc/9FTR-MCVY . 2024. url: https://news.samsung.com/us/ samsung-google-cloud-bring-generative-ai-to-samsung-galaxy-s24-series/

  138. [147]

    Samsung Galaxy S25 Series Sets the Standard of AI Phone as a True AI Companion

    Samsung. Samsung Galaxy S25 Series Sets the Standard of AI Phone as a True AI Companion . permalink: https://perma.cc/RJ7U- 28WN . 2025. url: https://news.samsung.com/global/ samsung-galaxy-s25-series-sets-the-standard-of-ai-phone-as-a-true-ai-companion

  139. [148]

    Samsung Knox Documentation: Knox Vault

    Samsung. Samsung Knox Documentation: Knox Vault . permalink: https://perma.cc/7QK3-UA9C

  140. [149]

    Saumsung Knox Documentation; Data processing for Galaxy AI

    Samsung. Saumsung Knox Documentation; Data processing for Galaxy AI . permalink: https:// perma.cc/6ERX-Y8TQ . 2025. url: https://docs.samsungknox.com/admin/knox-platform-for- enterprise / knox - service - plugin / configure - advanced - policies / data - processing - for - galaxy-ai/

  141. [150]

    Use features with Galaxy AI on your Galaxy phone and tablet

    Samsung. Use features with Galaxy AI on your Galaxy phone and tablet . permalink: https : / / perma.cc/E6YA-9ZWR. 2025. url: https://www.samsung.com/us/support/answer/ANS10000753/

  142. [151]

    What are the Galaxy AI Features on the Galaxy S25 Series? permalink: https://perma

    Samsung. What are the Galaxy AI Features on the Galaxy S25 Series? permalink: https://perma. cc/VD4H-P7G4. 2025. url: https://www.samsung.com/uk/mobile-phone-buying-guide/samsung- galaxy-s25-galaxy-ai-features/

  143. [152]

    What is the ‘Enhanced Data Protection ’ function, and when can I use it? permalink: https://perma.cc/2B37- KUBA

    Samsung. What is the ‘Enhanced Data Protection ’ function, and when can I use it? permalink: https://perma.cc/2B37- KUBA . 2023. url: https://www.samsung.com/ae/support/mobile- devices/what-is-the-enhanced-data-protection-function-and-when-can-i-use-it

  144. [153]

    SoK: Content Moderation for End-to-End Encryption

    Sarah Scheffler and Jonathan Mayer. SoK: Content Moderation for End-to-End Encryption . https: //arxiv.org/abs/2303.03979. 2023. arXiv: 2303.03979 [cs.CR]

  145. [154]

    WhatsApp outage results in a scramble for messaging alternatives

    Gadjo Sevilla. WhatsApp outage results in a scramble for messaging alternatives . https : / / www . emarketer . com / content / whatsapp - outage - results - scramble - messaging - alternatives. permalink: https://perma.cc/LN6D-8DQ4. 2022

  146. [155]

    Data Brokers Are a Threat to Democracy

    Justin Sherman. Data Brokers Are a Threat to Democracy . https : / / www . wired . com / story / opinion-data-brokers-are-a-threat-to-democracy/ . permalink: https://perma.cc/AH8Z- JAR7. 2021

  147. [156]

    Membership inference attacks against machine learning models

    Reza Shokri et al. “Membership inference attacks against machine learning models”. In: 2017 IEEE symposium on security and privacy (SP) . IEEE. 2017, pp. 3–18

  148. [157]

    Murky Consent: An Approach to the Fictions of Consent in Privacy Law

    Daniel J. Solove. “Murky Consent: An Approach to the Fictions of Consent in Privacy Law”. In: Boston University Law Review (2024). permalink: https://perma.cc/8ZGU-4M3F

  149. [158]

    Privacy Self-Management and the Consent Dilemma

    Daniel J. Solove. “Privacy Self-Management and the Consent Dilemma”. In: 126 Harvard Law Re- view 1880 (2013), GWU Legal Studies Research Paper No. 2012-141, GWU Law School Public Law Research Paper No. 2012-141 (2013). permalink: https://perma.cc/4THR-HS4F

  150. [159]

    McNealy Stephanie T

    Jasmine E. McNealy Stephanie T. Nguyen. I, Obscura — Illuminating deceptive design patterns in the wild. https://pacscenter.stanford.edu/wp-content/uploads/2021/07/I-Obscura-Zine.pdf . permalink: https://perma.cc/2UAD-ZAFN. 2021. 50

  151. [160]

    Data Protection

    European Data Protection Supervisor. Data Protection . https : / / www . edps . europa . eu / data - protection/data-protection_en. permalink: https://perma.cc/H8KV-SQXN

  152. [161]

    When the Terms of Service Change to Make Way for A.I

    Eli Tan. When the Terms of Service Change to Make Way for A.I. Training . https://www.nytimes. com/2024/06/26/technology/terms-service-ai-training.html . permalink: https://perma. cc/NJL3-U22V. 2024

  153. [162]

    End-to-End Encryption F AQ

    Telegram. End-to-End Encryption F AQ. https : / / tsf . telegram . org / manuals / e2ee - simple. permalink: https://perma.cc/68ST-YUFQ. (Visited on 01/15/2025)

  154. [163]

    The world’s most valuable resource is no longer oil, but data

    The Economist. The world’s most valuable resource is no longer oil, but data . https : / / www . economist.com/leaders/2017/05/06/the- worlds- most- valuable- resource- is- no- longer- oil-but-data . permalink: https://perma.cc/YF85-8QLS. 2017

  155. [164]

    AI Assistants in Encrypted Messaging: Moving Too Fast in the Dark? NYU Cen- ter for Data Science on Medium

    Stephen Thomas. AI Assistants in Encrypted Messaging: Moving Too Fast in the Dark? NYU Cen- ter for Data Science on Medium. Mar. 2025. url: https : / / nyudatascience . medium . com / ai - assistants-in-encrypted-messaging-moving-too-fast-in-the-dark-8680ab195834

  156. [165]

    SoK: Secure Messaging

    Nik Unger et al. “SoK: Secure Messaging”. In: IEEE S&P . 2015

  157. [166]

    L VI: Hijacking Transient Execution through Microarchitectural Load Value Injection

    Jo Van Bulck et al. “L VI: Hijacking Transient Execution through Microarchitectural Load Value Injection”. In: IEEE S&P . 2020

  158. [167]

    Better the Devil You Know: Exposing the Data Sharing Practices of Smart- phone Apps

    Max Van Kleek et al. “Better the Devil You Know: Exposing the Data Sharing Practices of Smart- phone Apps”. In: Proceedings of the 2017 CHI Conference on Human Factors in Computing Systems . 2017

  159. [168]

    SoK: SGX.Fail: How Stuff Gets eXposed

    Stephan Van Schaik et al. “SoK: SGX.Fail: How Stuff Gets eXposed”. In: IEEE S&P . 2024

  160. [169]

    Position: Will we run out of data? Limits of LLM scaling based on human- generated data

    Pablo Villalobos et al. “Position: Will we run out of data? Limits of LLM scaling based on human- generated data”. In: Forty-first International Conference on Machine Learning

  161. [170]

    Meta ‘pauses’ AI data collection in EU following Irish DPC request

    Adrian Weckler. Meta ‘pauses’ AI data collection in EU following Irish DPC request . https://www. independent.ie/business/technology/meta-pauses-ai-data-collection-in-eu-following- irish-dpc-request/a2000443736.html. permalink: https://perma.cc/XDG6-ZH5H. 2024

  162. [171]

    “Consent”

    Wex. “Consent”. In: Legal Information Institute (2022). permalink: https://perma.cc/RB7X-UL4A

  163. [172]

    WhatsApp Encryption Overview, v.8

    WhatsApp. “WhatsApp Encryption Overview, v.8”. In: (2024). permalink: https://perma.cc/ GQ5W-RUGX. url: https://whatsapp.com/security/WhatsApp-Security-Whitepaper.pdf

  164. [173]

    WhatsApp F AQ: About AIs from Meta

    WhatsApp. WhatsApp F AQ: About AIs from Meta . permaink: https : / / perma . cc / 8D6T - F3CQ

  165. [174]

    WhatsApp F AQ: About using AI on WhatsApp

    WhatsApp. WhatsApp F AQ: About using AI on WhatsApp . permalink: https://perma.cc/EG99- GA7B. 2025. url: https://faq.whatsapp.com/1002544104126998

  166. [175]

    Rethinking privacy in machine learning pipelines from an information flow control perspective

    Lukas Wutschitz et al. “Rethinking privacy in machine learning pipelines from an information flow control perspective”. In: arXiv preprint arXiv:2311.15792 (2023)

  167. [176]

    Controlled-Channel Attacks: Deterministic Side Channels for Untrusted Operating Systems

    Yuanzhong Xu, Weidong Cui, and Marcus Peinado. “Controlled-Channel Attacks: Deterministic Side Channels for Untrusted Operating Systems”. In: IEEE S&P . 2015

  168. [177]

    url: https://faq.whatsapp.com/2257017191175152

  169. [181]

    semantic

    Zoom. Security Guide: Zoom . https : / / explore . zoom . us / docs / doc / Zoom - Security - White - Paper.pdf. permalink: https://perma.cc/P4MV-KWLE. 2021. A E2EE Features Extended Evaluation Framework In the course of evaluating the many ways in which AI features might be i...

  170. [2019]

    arXiv: 1909.01066 [cs.CL]

  171. [2024]

    url: https://docs.samsungknox.com/admin/fundamentals/whitepaper/samsung- knox- for-android/core-platform-security/knox-vault/

  172. [2025]

    url: https://support.google.com/messages/answer/10252671

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.