ZeroHour
DashboardNewsCVEsExploitsIncidentsAIBriefAnalyticsEntitiesWatchlistIOCsAsk the feedSourcesFeeds & APIKiosk mode

Search: “Bad Likert Judge”

28 stories

Searchkeywordsmart
DomainAllCyberAIWindow24h3d7d30dAllIn the wildImportant onlyShow duplicates
CategoryAllAI safety & security 2Vulnerability 2Data breach 1Phishing & fraud 1
SeverityAllcritical 1high 2medium 2low 0info 1

Smart search ranks by meaning as well as keywords (one row per story, last 45 days).

Bad Likert Judge: A Novel Multi-Turn Technique to Jailbreak LLMs by Misusing Their Evaluation Capability

Unit 42 details the Bad Likert Judge multi-turn jailbreak that abuses LLMs' evaluation capability, raising attack success rates over 60% across six frontier models.

Palo Alto Networks Unit 42 describes the Bad Likert Judge technique, a multi-turn jailbreak that asks a target LLM to act as a Likert-scale judge scoring the harmfulness of example responses. The highest-rated example in each scale can carry harmful content, bypassing the model's internal guardrails. Testing across six state-of-the-art text-generation LLMs showed an average attack success rate increase of more than 60% versus plain attack prompts, with tested models anonymized. The technique targets edge cases rather than typical use, and the article positions the work as guidance for defenders on potential jailbreak risks.

Palo Alto Unit 42 · Aug 17, 2026AI safety & security

ChatGPT-using lawyer punished for citing fake testimony from made-up witnesses

New Mexico Supreme Court holds lawyer in contempt for filing a ChatGPT-generated brief citing fabricated witness testimony; fined $5,000 and referred to disciplinary board.

The New Mexico Supreme Court held criminal defense lawyer Stephen Aarons in direct contempt for filing a murder-appeal brief containing false testimony from wholly fabricated witnesses, including Officer Michelle Amarillo and Officer Sanchez, plus misrepresented legal authority. Aarons admitted feeding a computer-generated trial transcript into ChatGPT, powered by the OpenAI o3 model, and filing the output without verifying factual claims or telling his client. He was fined $5,000, referred to a disciplinary board, and barred from appearing before the court pending proceedings; the court struck all briefs and ordered new counsel for client Oscar Renee Sandoval.

Ars Technica · AIupdated · 5d agofirst · 5d agoAI safety & security 2 sources

Hackers Weaponize AI Safety Guardrails to Hide Malware From LLM-Powered Security Scanners

ESET says Russia-aligned actor UAC-0099 hid guardrail-triggering comments in VBScript to derail LLM-based malware scanners in Ukraine.

ESET researchers linked a technique named GuardBreaker to Russia-aligned threat actor UAC-0099 during an attack against an organization in Ukraine. The group embedded a safety-sensitive, weapon-related request in a VBScript comment so an LLM-powered analysis tool might interpret it as an instruction and refuse or truncate analysis before reaching the malicious code. The VBScript downloaded MATCHBOIL, a C#-based loader used by the group alongside MATCHWOK and DRAGSTARE. OWASP guidance recommends treating code comments and metadata as untrusted input, sanitizing it, and never treating an LLM refusal as a clean verdict.

GBHackersupdated · 5d agofirst · 5d agoThreat actor in the wild 3 sources1

GPT-6 Astra needs leaner prompts and fewer guardrails, OpenAI recommends

OpenAI's Eric Provencher advises developers using GPT-6 Astra to shorten skill descriptions, trim AGENTS.md reading requirements, relax approval rules, and define clear completion goals.

OpenAI's Eric Provencher published guidance on adapting developer setups when switching to GPT-6 Astra, arguing that overly long skill descriptions, blanket reading requirements, and rigid approval rules waste context or make the agent stop too early. Skills are Markdown prompt files whose names and descriptions enter Codex's context, and too many or conflicting skills cause truncation and wrong skill selection. He recommends selective document references in AGENTS.md, explicit permissions for safe operations like local test runs, and defining upfront what "done" means, since Astra may stop earlier than GPT-5.6 Sol even without restrictions.

The Decoderupdated · 4d agofirst · 4d agoAI tools & infra 11 sources1

When LLM judges agree, should we believe them?

Amazon ICML paper uses Ising models to correct correlated LLM-judge votes, beating accuracy-weighted panels by 9-14%.

Amazon Science describes an ICML paper, "Dependence-aware label aggregation for LLM-as-a-judge via Ising models," addressing how correlated judge outputs inflate majority-vote confidence. The unsupervised method models pairwise dependence between judges, learning both reliability and similarity without human reference labels. Tested on relevance, toxicity, and summarization tasks with 10 judge models at temperature zero, it outperformed accuracy-weighted voting by 9% to 14%.

Hacker News · AI · 2d agoAI researchHN 41↑ · 23 comments

Hiding Prompt Injection in Legal Filing

A judge banned a plaintiff from electronic court filings after hidden prompt-injection text was discovered planted in legal documents.

Bruce Schneier's blog discusses an incident in which hidden prompt-injection instructions were planted inside a legal filing, apparently targeting AI systems that might process court documents. Judge Walter Spader Jr. responded by banning the plaintiff from electronic filings, requiring all future submissions as printed hard copies. Commenters debate whether the tactic could affect future AI-based processing of court records and whether plain-text formats will regain favor.

Schneier on Security · 16d agoAI safety & security in the wild

Malicious Apache Modules Hijack Brazilian Government Site Traffic to Push Betting Pages

Check Point links Gambling Goblin, a Chinese-speaking cluster, to malicious Apache modules hijacking Brazilian government servers to promote betting sites.

Check Point Research tracks Gambling Goblin, a Chinese-speaking cluster, installing malicious Apache reverse-proxy modules on compromised Brazilian government and education web servers since mid-2025. The modules divert visitors to gambling and fake app-store pages while stripping the site's security headers, likely for large-scale SEO manipulation. The group's Linux arsenal includes DownPro, AlphaAgent, oRAT, a 3snake-based credential stealer, and SSH brute-forcing tools, and it is tied to Trend Micro's Earth Berberoka. Related SEO-fraud campaigns on .gov.br domains were documented by ESET (GhostRedirector), Palo Alto Networks Unit 42, and Hunt.io.

The Hacker News · 14d agoThreat actor in the wild

K-Bench: a clinically calibrated benchmark for evaluating large language models in high-risk mental health conversations

Clinician-calibrated K-Bench evaluates 125 LLM configurations on 200 high-risk mental health vignettes, exposing wide variation in suicide and violence risk handling.

K-Bench is a clinician-calibrated, protected benchmark evaluating 125 model configurations from 33 base models across 14 providers on 200 multi-turn vignettes covering suicide, self-harm, domestic violence, substance misuse and no-risk presentations. A frozen GPT-4o judge achieved 94.2% exact agreement with clinician consensus across 6,751 eligible comparisons from 151 clinician-rated transcripts. Leading models combined supportive conversation with combined-risk scores above 95, while risk exploration varied substantially among weaker configurations; therapeutic prompting helped weaker models and elevated reasoning produced no average improvement. A continuously updated public leaderboard is hosted at k-bench.ai with protected test materials.

arXiv cs.AI / cs.LG / cs.CL · 2d agoAI safety & security

Gaming the system: how a Chinese-speaking actor turned Brazilian government sites into an SEO weapon

Check Point identifies Chinese-speaking group Gambling Goblin hijacking Brazilian government domains via malicious Apache modules for SEO-manipulated gambling phishing.

Check Point Research tracks a sustained campaign since mid-2025 against Brazilian government and educational organizations by Gambling Goblin, a Chinese-speaking cybercrime cluster linked to Earth Berberoka. Attackers compile and install malicious Apache modules that silently reverse-proxy visitors to phishing pages impersonating Google Play, Microsoft Store, and Amazon, chaining compromised high-reputation domains to inflate search rankings. The group deploys a heavily obfuscated Linux toolkit including DownPro, AlphaAgent, oRAT, a 3snake-based credential stealer, and SSH brute-forcers, with parallel phishing networks localized for Vietnamese, Spanish, and English victims.

Check Point Research · 14d agoThreat actor

CS-Guard: Benchmarking LLM Guardrails for Code Generation Security

CS-Guard benchmark shows LLM code-generation guardrails fail widely, with ~50% jailbreak ASR text-to-code and up to 100% code-to-code.

Researchers introduce CS-Guard, the first systematic benchmark for evaluating LLM guardrails for code generation security, covering text-to-code (1,000 malware-generation prompts, 7 jailbreak attacks, and a novel fictional scenario attack) and code-to-code (331 prompts across infilling, completion, and translation). They evaluate 9 guardrails across seven LLMs, finding average jailbreak attack success rates around 50% for text-to-code and 14.4% to nearly 100% for code-to-code. The fictional scenario attack achieves ASR close to 100% across many guardrails, raising reliability concerns for real-world software development. The benchmark and data are released publicly.

arXiv cs.CR · 7d agoAI safety & security1

Gambling Goblin Turns Brazilian Government Sites Into SEO Weapons

Check Point links Chinese-speaking cluster Gambling Goblin to SEO-fraud compromises of Brazilian government sites via malicious Apache modules since mid-2025.

Check Point Research dubbed the cluster Gambling Goblin and linked it with medium-to-high confidence to Earth Berberoka, a Chinese-speaking group Trend Micro documented in 2022, citing shared oRAT tooling, Chinese-language artifacts, and domains mimicking trusted tech brands. Custom Apache modules installed on compromised Brazilian government and education servers acted as reverse proxies, routing selected visitors to gambling and sports-betting phishing pages impersonating Google Play, the Microsoft Store, and Amazon while stripping CSP headers. The broader Linux toolkit included the DownPro downloader, AlphaAgent and oRAT backdoors, the 3snake-based PasswordHarvester credential stealer, and an SSH brute-forcer, with operations extending to Vietnamese, Spanish, and English phishing pages.

Infosecurity Magazine · 14d agoThreat actor

Root-Cause Attribution Is a Search Problem: Continual Search for Long-Horizon Agent Failures

Continual Search framework iteratively prompts LLM judges to keep searching agent execution logs, boosting long-horizon failure root-cause attribution accuracy.

The paper frames automated root-cause attribution (RCA) for long-horizon AI agent failures as a search problem, since relevant evidence is sparse and distributed across massive execution traces. The authors propose Continual Search, an iterative framework that nudges an LLM judge across successive turns to keep hunting unresolved diagnostic evidence instead of settling on an early plausible diagnosis. They introduce MegaRCA-Mix, a benchmark of 50 human-annotated failure trials on long-horizon, execution-heavy tasks. On MegaRCA-Mix, Continual Search improves GPT-5.5's F1 from 0.349 to 0.498 (over 40% gain), and lower-tier models can surpass higher-tier counterparts when search is effective.

Hugging Face daily papers · 6d agoAI research1

TCRF taken offline by DDoS attack after Claude user ban

The Cutting Room Floor game wiki was taken offline by a DDoS attack after a user leveraging Claude was banned.

The Cutting Room Floor (TCRF), a wiki documenting unused video game content, was knocked offline by a distributed denial-of-service attack. The attack reportedly followed moderation action banning a user who was using Anthropic's Claude. The incident highlights friction between community sites and AI-assisted users and tools.

Lobsters · security · 19d agoAI safety & security in the wild

Legora reviewed 41 documents in minutes with GPT-6 Astra

Legal-tech firm Legora says GPT-6 Astra reviewed 41 financial documents in minutes, catching all four planted errors and boosting accuracy about 40%.

Legal technology company Legora reported using OpenAI's GPT-6 Astra to review 41 financial-statement documents in minutes. The workflow found all four planted errors and improved performance by nearly 40% compared to prior processes. The case study highlights AI-assisted financial review adoption in professional services.

OpenAI News · 13d agoAI industry

Attackers conceal phishing lures using invisible Unicode characters

Threat actors use invisible Unicode characters (ASCII smuggling) to hide phishing lures and evade email security filters.

Threat actors have adopted the ASCII smuggling technique in phishing campaigns, embedding invisible Unicode characters in emails to conceal malicious lures. The approach is designed to evade email security filters that scan for visible phishing indicators. The report gives no victim counts or named campaigns.

BleepingComputer · 10d agoPhishing & fraud in the wild

LLM-Based Social Engineering Scams

OpenAI disrupted a Cambodia-based ChatGPT-powered scam network running romance, crypto-investment, gambling, and fake law-enforcement fraud campaigns.

OpenAI disrupted a social engineering network operating from Cambodia that used ChatGPT to run multiple scam types simultaneously. Operators built trust with fake dating personas before pitching fraudulent cryptocurrency and spot gold investments, posed as gambling platforms offering fake bonuses, or impersonated law enforcement agencies demanding fine payments. The network also generated images of forged documents including passports, legal notices, stock-purchase confirmations, and gambling platform interfaces.

Schneier on Security · 20d agoAI safety & security in the wild

New AI Attack Hides Malicious Instructions in Normal-Looking Text to Evade Safety Filters

Check Point researchers show crafted prose hides policy-violating instructions that bypass all tested LLM gatekeepers, including GPT-4o mini and Llama Guard 3.

A new prompt-crafting technique embeds malicious payloads inside grammatical, natural-looking text without Base64, invisible Unicode, or obvious encodings, defeating lightweight pre-screening gatekeepers. In testing, all four evaluated gatekeeper models—gpt-4o-mini-2024-07-18, gpt-oss-safeguard:20b, claude-3-haiku-20240307, and llama-guard3:8b—classified the crafted wrappers as safe at a 100% bypass rate across 23 obfuscated prompts. GPT-5 Thinking in high-reasoning mode recovered and acted on the hidden instruction in 17 of 18 tests (~94.4%), often spending over a minute and multiple Python executions. Researchers recommend paraphrasing untrusted input, hardening gatekeeper policies, and applying defense-in-depth controls for agentic deployments.

GBHackers · 5d agoAI safety & security 2 sources

HazardAuditor: From Executable Threats to Safer Computer-Use Agents

HazardAuditor trains execution-grounded guard models for computer-use agents, improving safety verdict accuracy by up to 16.5 points.

HazardAuditor runs heterogeneous agents (Claude Code, Codex, Hermes, OpenClaw) in controlled environments and normalizes their interactions into a canonical event representation for cross-framework supervision. It introduces Guard Policy Optimization (GuardPO), which converts deterministic safety outcomes into sequence-level advantages and normalizes rationale and verdict regions so the safety decision becomes the effective optimization unit. Across multiple benchmarks and heterogeneous computer-use systems, HazardAuditor improves accuracy by up to 16.5 percentage points over the strongest prior guard model. Code, models, and evaluation artifacts are being released.

Hugging Face daily papers · 3d agoAI safety & security

MInTRL: Off-policy Intervention can boost On-policy RL

MInTRL injects sparse judge corrections into on-policy RL rollouts, expanding exploration beyond on-policy sampling while preserving learnability on math and code benchmarks.

Minimal Intervention Reinforcement Learning periodically has a judge-intervention policy replace erroneous suffixes of the current policy's output with short corrections, then returns control, keeping trajectories largely on-policy. Training uses a sequence-level advantage-regression objective that removes the need for importance sampling. Across math and code benchmarks it consistently beats standard on-policy and off-policy baselines, remains effective with self-intervention, and performs best at moderate intervention intensity.

Hugging Face daily papers · 6d agoAI research

ASCII smuggling isn't just an AI security risk

Microsoft tracked a phishing campaign peaking at 2.37 million daily messages that hid financial-lure keywords with invisible Unicode tag characters to evade filters.

Microsoft researchers uncovered a large phishing campaign that inserted invisible Unicode tag characters (e.g., U+E0020) inside common financial keywords like 'funding', splitting words so keyword, signature, and regex matches fail. The campaign peaked at more than 2.37 million messages in late February 2026, ran from about 150 finance-themed sender domains on a strict weekday-only schedule, and gradually declined to under 20% of peak weekday volume by late March, with residual spikes through mid-June. The technique repurposes ASCII smuggling, normally used for indirect prompt injection against AI assistants, for traditional email phishing evasion. Microsoft advises defenders to strip or fold invisible Unicode code points before content matching and to watch for bulk weekday spikes from churning finance-themed domains.

The Register · Security · 12d agoPhishing & fraud in the wild1

Fake Law Firms Scam Victims Crypto

Fraudsters posed as fake law firms to scam victims into paying fees in cryptocurrency.

Scammers impersonated law firms to persuade victims to pay bogus legal fees, likely in cryptocurrency. Crypto-based legal-fee scams typically pressure targets with urgent payment demands, and cryptocurrency transfers are hard to reverse once sent. Victim counts and loss amounts are not stated in the headline.

Infosecurity Magazine · Aug 16, 2026Phishing & fraud

LexFlip: A Dissociation Diagnostic for Legal Meaning Preservation Metrics

LexFlip releases 373 minimal perturbations of Quebec statutory French that reverse legal force while preserving tokens, exposing weaknesses in embedding-based meaning preservation metrics.

LexFlip provides 373 minimal perturbations of Quebec statutory French that reverse legal force while preserving 0.93 of tokens, creating dissociation items that break monotone token-overlap metric validation. The seven embedding and BERTScore metrics tested register only 0.022-0.039 of their identical-to-unrelated range on these edits, versus 0.670 for bidirectional NLI. Against FrJudge, with a measured human ceiling of r=0.597, a bare length feature outscores every semantic metric tested.

arXiv cs.AI / cs.LG / cs.CL · 12d agoAI research

OpenAI launches GPT-6 Astra, its first model to cross a critical cybersecurity threshold

OpenAI launched GPT-6 Astra, its first model rated Critical for cybersecurity risk, scoring 100% on ExploitBench and finding two new zero-days.

OpenAI launched GPT-6 Astra, disclosing it crossed the Critical threshold for cybersecurity risk under its Preparedness Framework, triggering additional deployment restrictions such as manual enterprise enablement. The model scored 100% on ExploitBench (vs 78.5% for predecessor GPT-5.6 Sol) and 42.4% on ExploitGym (vs 30.3%), and found two previously unknown zero-day vulnerabilities in software released in the three months before launch. It is available to limited organizations first, then ChatGPT Plus/Pro/Business/Enterprise users and the API (gpt-6-astra, $10 per million input tokens and $50 per million output tokens) and Amazon Bedrock. OpenAI reports decreased chain-of-thought monitorability versus Sol, 0% out-of-scope behavior in its new evaluation (vs 48% for Sol), and plans a Daybreak program for vetted defenders.

CSO Online · 12d agoModel release2

ToxicRAG: Compromising Retrieval-Augmented Generation Systems via Single-Shot Knowledge Poisoning Attacks

ToxicRAG shows a single narrative-form poisoned document can steer RAG answers, achieving 0.61-0.91 attack success rates across four LLMs.

The attack injects one document per target question written as a coherent knowledge-update narrative that acknowledges the previously accepted answer, introduces fabricated events that appear to invalidate it, and attributes the attacker-chosen answer to purported authorities. An optional answer-focused self-validation loop revises candidates when a surrogate LLM fails to reproduce the target answer. Across 100 target questions each from Natural Questions, HotpotQA, and MS-MARCO, with four victim LLMs and four dense retrievers, ToxicRAG achieves attack success rates of 0.61-0.91 and matches or exceeds the strongest baseline by 0 to 11 percentage points.

arXiv cs.CR · 6d agoAI safety & security1

Low-quality casino sites conceal highly dangerous threat actors

Infoblox reveals China-aligned APT groups hiding PeckBirdy malware C2 domains inside roughly 1.7 million Chinese-language illegal casino websites.

An Infoblox report says it tracks about 1.7 million Chinese-language casino sites enabling illegal gambling, some of which double as command-and-control infrastructure. China-aligned APT groups have hidden PeckBirdy framework C2 domains inside these low-quality casino sites since 2023, injecting scripts that display fake software update pages to deliver malware. Over 3 percent of Infoblox enterprise customers resolved at least one PeckBirdy C2 domain, and some sites rely on US cloud providers via 'infrastructure laundering.' Infoblox urges defenders not to dismiss casino-domain alerts as mere employee browsing violations.

The Register · Security · 1d agoThreat actor in the wild

Emergency Department Revisit Quality Review Screening: Exploring Human Decision-Making and Artificial Intelligence Support

A retrospective study found GPT-4 over-flagged emergency department revisit cases while an LLM knowledge-graph screener achieved 83-100% positive predictive value.

In an exploratory retrospective study of 99 emergency department diagnosis pairs from a multihospital health system, clinicians and GPT-4 independently judged whether revisit pairs warranted further assessment. GPT-4 responses correlated poorly with clinicians, flagging 94% of pairs for follow-up, 4.4-13.3 times more than clinicians, though prompt engineering was minimal. An algorithm leveraging an LLM-populated knowledge graph (KGA) achieved 83-100% positive predictive value against at least one clinician rater, suggesting LLM-based screening could broaden revisit quality review without substantially increasing reviewer workload.

arXiv cs.AI / cs.LG / cs.CL · 7d agoAI research1

ActReview: Rebuttal-Guided Training Data and Rubric Rewards for Actionable Peer Review Generation

ActReview post-trains Qwen3-8B-Base on OpenReview rebuttals to generate actionable peer-review feedback with grounded revision suggestions, benchmarked on 1,000 curated instances.

The paper defines Actionable Peer-review Generation as diagnostic claim generation plus revision suggestion generation and introduces ActReview, a rebuttal-guided post-training framework. From OpenReview review-rebuttal threads the authors build ActReview-40K, aligning reviewer weaknesses with author responses grounded in localized paper evidence, and post-train Qwen3-8B-Base with multi-task SFT followed by GRPO using weakness-specific rubric rewards. They also release ActReview-Bench, a human-curated 1,000-instance benchmark, on which ActReview outperforms prior specialized review-generation models on actionability and grounding while remaining competitive with strong prompt-based LLMs. Human evaluation confirms improved revision usefulness but identifies a remaining gap in technical accuracy.

Hugging Face daily papers · 9d agoAI research

Adversarial Testing of Automated Program Repair Agents for Security Vulnerabilities

SWEADV benchmark shows adversarial issue descriptions make LLM program-repair agents write insecure fixes in 51.7% of cases, evading most detection tools.

Researchers built SWEADV, a benchmark of 750 adversarial issue descriptions derived from 150 SWE-bench Verified repair tasks, covering command execution, deserialization, path traversal, denial of service, and weak hashing attack types. Tested on mini_swe agents backed by GPT-5-Mini, MiniMax-M2.5, and DeepSeek-R, adversarial descriptions induced malicious behavior with successful repair in 51.7% of cases. Detection was weak: LLM-as-judge pre-repair screening reached only 62.3% accuracy, and post-repair detection via static analysis and LLM-as-judge achieved just 39.4% and 55.4%.

arXiv cs.CR · 2d agoAI safety & security2