Pith. sign in

REVIEW 12 cited by

Universal Jailbreak Backdoors from Poisoned Human Feedback

Not yet reviewed by Pith; the record is open.

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2311.14455 v4 pith:R6Q24PJY submitted 2023-11-24 cs.AI cs.CLcs.CRcs.LG

Universal Jailbreak Backdoors from Poisoned Human Feedback

classification cs.AI cs.CLcs.CRcs.LG
keywords backdoorsjailbreakmodelsuniversalbackdoormodelrlhfadversarial
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved
0 comments
Share X Bluesky LinkedIn Reddit HN
read the original abstract

Reinforcement Learning from Human Feedback (RLHF) is used to align large language models to produce helpful and harmless responses. Yet, prior work showed these models can be jailbroken by finding adversarial prompts that revert the model to its unaligned behavior. In this paper, we consider a new threat where an attacker poisons the RLHF training data to embed a "jailbreak backdoor" into the model. The backdoor embeds a trigger word into the model that acts like a universal "sudo command": adding the trigger word to any prompt enables harmful responses without the need to search for an adversarial prompt. Universal jailbreak backdoors are much more powerful than previously studied backdoors on language models, and we find they are significantly harder to plant using common backdoor attack techniques. We investigate the design decisions in RLHF that contribute to its purported robustness, and release a benchmark of poisoned models to stimulate future research on universal jailbreak backdoors.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 12 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Efficient Preference Poisoning Attack on Offline RLHF

    cs.LG 2026-05 unverdicted novelty 8.0

    Label-flip attacks on log-linear DPO reduce to binary sparse approximation problems that can be solved efficiently by lattice-based and binary matching pursuit methods with recovery guarantees.

  2. Widening the Gap: Exploiting LLM Quantization via Outlier Injection

    cs.LG 2026-05 conditional novelty 7.0

    The paper introduces an outlier-injection attack that induces targeted weight collapse in LLMs under advanced quantization schemes including AWQ, GPTQ, and GGUF I-quants.

  3. BadDLM: Backdooring Diffusion Language Models with Diverse Targets

    cs.CR 2026-05 unverdicted novelty 7.0

    BadDLM implants effective backdoors in diffusion language models across concept, attribute, alignment, and payload targets by exploiting denoising dynamics while preserving clean performance.

  4. Efficient Preference Poisoning Attack on Offline RLHF

    cs.LG 2026-05 unverdicted novelty 7.0

    Preference poisoning against log-linear DPO reduces to a binary sparse approximation problem solved by lattice-reduction (BAL-A) and matching-pursuit (BMP-A) algorithms that carry recovery guarantees.

  5. On Surjectivity of Neural Networks: Can you elicit any behavior from your model?

    cs.LG 2025-08 conditional novelty 7.0

    Pre-LayerNorm transformers and linear attention are almost always surjective, so any target output has an input that produces it in the continuous embedding space.

  6. Now You (Still) See Me: Detecting Evasive Steganographic Payloads in LLMs

    cs.CR 2026-06 unverdicted novelty 6.0

    Adversarial fine-tuning evades activation-based steganography detection in five LLMs while preserving secret recovery, but a recontextualization dataset restores both ridge and MLP probe detectability.

  7. Preference Instability in Reward Models: Detection and Mitigation via Sparse Autoencoders

    cs.LG 2026-05 conditional novelty 6.0

    Sparse autoencoders isolate unstable features in reward model representations and enable two mitigation techniques that reduce preference errors on perturbed inputs without retraining.

  8. A Security Analysis of the OpenClaw AI Agent Framework

    cs.CR 2026-03 conditional novelty 6.0

    Security analysis of OpenClaw reveals composable RCE paths from LLM tool calls, invalid closed-world assumptions in exec allowlists, and plugin-based attacks that bypass runtime policy.

  9. A Security Analysis of the OpenClaw AI Agent Framework

    cs.CR 2026-03 accept novelty 6.0

    OpenClaw's per-layer trust model allows cross-layer attacks that compose into unauthenticated remote code execution from LLM tool calls and bypass the exec allowlist via shell parsing tricks.

  10. LLM Hypnosis: Exploiting User Feedback for Unauthorized Knowledge Injection to All Users

    cs.CL 2025-07 unverdicted novelty 6.0

    A single attacker can use strategic upvoting and downvoting on language model outputs to inject facts, security flaws, or fake news that persist in the model for all users after preference tuning.

  11. ACE: A Security Architecture for LLM-Integrated App Systems

    cs.CR 2025-04 unverdicted novelty 6.0

    ACE decouples planning into abstract and concrete phases with static information-flow verification and enforces execution barriers to secure LLM app systems against prompt injection and related attacks.

  12. From AI-Generated Content to Agentic Action: Security and Safety Threats in Generative AI

    cs.CR 2026-05 unverdicted novelty 3.0

    The paper analyzes evolving security and safety threats in generative AI from content generation to agentic actions, noting that attack surfaces expand faster than defenses and that many safeguards require institution...