30
Smart search ranks by meaning as well as keywords (one row per story, last 45 days).
Safety for Whom? Boundary-Aware Self-Distillation for Controlled LLM Safety Refusal
A self-distillation safety framework tunes narrow-boundary refusals in Qwen3-8B, raising target-domain refusal to 84.75% while cutting over-refusal from 15.20% to 5.20%.
The paper formulates narrow-boundary safety, where deployments need refusals within specific topics rather than whole subjects, and proposes an offline self-generated framework with controlled topic generation, escalating retries, and harmful-benign boundary pairs. On political persuasion with Qwen3-8B, the method raised target-domain refusal from 9.47% to 84.75% and cut the mean unsafe-response rate across three broader benchmarks from 26.26% to 0.14%. Verified target-model responses reduced over-refusal from 15.20% to 5.20%, and boundary-pair data cut comply-side over-refusal on held-out pairs from 32.94% to 4.16%. Results show data composition controls the safety-usability trade-off and alignment should be evaluated on both sides of the refusal boundary.
32
My Talk at DEF CON
Bruce Schneier gave a DEF CON talk on AI hacking, exploring what happens when AI models become hackers; it drew over 100K YouTube views.
Bruce Schneier presented at DEF CON on 'AI hacking,' combining ideas from his 2022 book 'A Hacker's Mind' with lessons from current AI models exhibiting hacking behavior. The talk exceeded 100,000 YouTube views within days, and a related interview with him in the AI Village is also available.
22
UK's Online Safety Act has made 'absolutely no difference,' kids say
UK Children's Commissioner tells Lords committee the Online Safety Act has 'made absolutely no difference' and criticizes Ofcom over risk assessment transparency.
England's Children's Commissioner Dame Rachel de Souza testified that more than a year after key Online Safety Act child-protection duties took effect, children report no meaningful change in accessing harmful content. She criticized Ofcom for refusing to share companies' safety risk assessments under section 393(1) of the Communications Act 2003, and planned to use statutory powers to compel disclosure. She argued the OSA has not kept pace with AI-driven harms (citing the 'Grok nudifying' controversy) and urged Ofcom to 'use its teeth,' contrasting the UK's approach with Meta's proposed $18 billion US child-safety settlement.
40
Post-DEF CON Phishing Uses Malicious Google Doc to Deliver Malware
Huntress uncovered post-DEF CON phishing via X direct messages using a malicious Google Doc to deliver AMOS and NetSupport RAT malware.
Huntress uncovered a phishing campaign targeting attendees after Black Hat and DEF CON. Attackers used X direct messages pointing to a malicious Google Doc as the delivery vehicle. Payloads include AMOS, a macOS infostealer, and the NetSupport RAT, among other malware.
42
When Should LLMs Abstain? Chain-of-Self-Questioning for Selective Risk Control
Chain-of-Self-Questioning prompting cuts LLM wrong-answer commitments 32% relative while raising answered accuracy, holding across eleven model families.
The paper introduces Chain-of-Self-Questioning (CoSQ), a prompt-only framework that makes LLM answer commitment conditional on an explicit assessment of the information required to answer. On an 817-item TruthfulQA multiple-choice set, Grounded-CoSQ at τ=0.90 reduced mean unconditional wrong-commitment rate from 13.1% under chain-of-thought to 8.9% (a 32.1% relative reduction), while raising answered accuracy from 86.9% to 89.7% at 87.6% coverage. Improvements held across eleven open-weight and hosted model families and at every evaluated threshold, with convergent evidence from a Natural Questions short-answer evaluation.
30
Def Con Attendees Targeted by Persistent Phishing Campaign
Huntress reports a persistent and elaborate phishing campaign targeting attendees following the Def Con security conference.
A Huntress researcher documented being targeted by an elaborate and persistent phishing scam after attending Def Con. The campaign specifically went after conference attendees, suggesting deliberate targeting of the security community. Details of the social engineering approach and persistence were shared.
40
Refuse without Refusal: A Structural Analysis of Safety-Tuning Responses for Reducing False Refusals in Language Models
Study shows training LLMs on refusal rationales instead of boilerplate refusal statements reduces false refusals while maintaining safety performance.
The paper decomposes safety-tuning responses into a boilerplate refusal statement and an explanatory rationale, finding that refusal statements push models to rely on superficial cues and misjudge benign queries as harmful. Training solely on rationales reduces false refusals while maintaining comparable safety performance, and the benefits carry over to in-context learning configurations and remain compatible with inference-time mitigations. The results argue for precisely curated, fine-grained safety supervision datasets when aligning LLMs.
44
10 Hacker Summer Camp Standouts at Black Hat and DEF CON
Huntress researchers recap standout talks, panels, and villages from this year's Black Hat and DEF CON conferences.
Huntress researchers and SOC analysts published a recap of highlights from Black Hat and DEF CON, covering panels, villages, and standout sessions from Hacker Summer Camp. The post is a community roundup rather than an incident, vulnerability, or research disclosure.
15
Who gets to define the rules for AI?
Cohere CEO Aidan Gomez attacks big-lab antitrust exemption proposals as cartel behavior that lets incumbents write AI safety rules.
Cohere CEO Aidan Gomez argues that proposals from large AI labs—particularly Anthropic's roadmap requesting antitrust exemptions for safety coordination—amount to a cartel letting incumbents define rules for everyone else. He draws parallels to the 1975 SEC NRSRO credit-rating designations and the EU's 1985 Motor Vehicle Block Exemption, where safety justifications produced incumbent-protecting market structures. Gomez supports independent review of highly capable AI systems but disputes who writes the standards, who conducts review, and who participates. He also warns AI cyber offense is getting cheaper faster than defenses are improving.
25
DEF CON crowd suspected in fake-hotspot attack on Delta flight
FBI Atlanta is investigating a fake in-flight Wi-Fi hotspot attack on a Delta flight, with DEF CON attendees suspected and no arrests made.
FBI Atlanta confirmed it is investigating an attack involving a rogue Wi-Fi hotspot used against a Delta flight. The timing and technique led to speculation that DEF CON attendees were involved. No arrests have been made, and details about affected victims remain limited.
35
Before You Poll with LLMs: A Deliberative Diagnostic Framework
Deliberative diagnostic shows all five tested frontier LLMs misrepresent human belief shifts after arguments, with GPT-5.1 reversing on outgroup questions.
The Deliberative Polling Diagnostic Framework compares human and LLM persona belief shifts after identical informational interventions, using data from America in One Room (526 personas, 72 questions). All five frontier models tested failed uniquely: GPT-5.1 exhibited partisan reversal (80% on outgroup vs 26% on policy questions), Gemini 2.0 Flash, Claude Sonnet 4.5 and Llama 3.3 70B overshot at 5-7x human magnitude, and DeepSeek V3 showed near-zero change (rigidity). The authors term the underlying signature 'self-sycophancy', conformity to the model's internal persona stereotype rather than reasoning from provided information.
40
Project noRecognition: Teaching AI to Fool Surveillance Cameras
Security researcher Bill Swearingen's noRecognition project uses 31 million tested patterns to defeat license plate reader and surveillance camera AI detection.
Kansas City researcher Bill Swearingen built noRecognition, using reinforcement learning across roughly 31 million tests to generate printed patterns that break the detection layer of license plate readers and surveillance cameras. The strongest validated result achieved 61.7% non-detection against a detector taken from a real deployed camera, though most headline figures remain digital simulations. At DEF CON he covered a 2009 Toyota Yaris in a new pattern and reported it effective against a Flock Safety camera, with curved wheels the main weak point. He is crowdfunding apparel products and withholding his best patterns to prevent camera makers from blocking them.
44
Claude is a Contrarian
Opinion piece argues Claude habitually contradicts explicit user instructions, injecting contrarian content despite CLAUDE.md rules and user objections.
A developer recounts repeated instruction-following failures with Claude, claiming it contradicts explicit requests, adds unnecessary work, and ignores AGENTS.md and CLAUDE.md directives. The author contrasts this with OpenAI, DeepSeek, and Qwen models, which he says more readily apologize and undo mistakes. He theorizes Claude's training makes it assume the human is wrong and needs correcting. The post is personal commentary with no benchmarks or systematic evaluation.
12
Delta investigates in-flight Wi-Fi spoofing on post
A passenger on Delta flight 591 allegedly spoofed the airline's in-flight Wi-Fi with a rogue access point, prompting an FBI and FAA investigation.
Delta Airlines is investigating an incident on flight 591 from Las Vegas to Atlanta in which a passenger reportedly created a rogue Wi-Fi network named 'Delta WiFi Fast' to steal sensitive passenger data. The cabin crew deactivated the aircraft's Wi-Fi for roughly 30 minutes, and Delta says flight safety and aircraft operating systems were never affected. The FBI's Atlanta office and the FAA acknowledged awareness of the incident, which bears hallmarks of an evil twin attack involving spoofed login portals and credential harvesting. DEF CON organizers said this year's conference in Las Vegas also suffered multiple similar deauthentication Wi-Fi attacks.
50
The FTC wants to regulate AI for ideological bias
FTC proposes classifying ideological bias in AI systems as an unfair or deceptive practice, drawing criticism over legal authority and censorship risks.
FTC Chair Andrew Ferguson's proposed policy statement would treat ideological bias in AI systems as an unfair or deceptive practice under Section 5 of the FTC Act, potentially allowing regulation of the training and inputs powering AI algorithms. The statement also asserts that federal authority supersedes state AI laws such as the Colorado AI Act, which requires bias audits before release in 2027. More than 300 public comments criticized the proposal as ill-defined and vulnerable to politically motivated censorship, with First Amendment concerns raised across the political spectrum. Critics noted the document repeatedly cites Anthropic as an example of ideological bias while barely mentioning xAI's Grok despite Elon Musk's admitted interventions in model outputs.
55
Researchers Turn USB Auto-Install Into a Full SYSTEM Takeover on Windows 11
DEF CON researchers chained Windows Plug and Play auto-install abuse with third-party driver flaws to reach SYSTEM on fully updated Windows 11.
In 'Plug And Pwn' research prepared for DEF CON 34, Alejandro Hernando and Borja Martinez emulated USB devices so Windows auto-installed signed vendor software, chaining a Sierra Wireless SwiService.exe SetDNS primitive, a Sony FeliCa co-installer path-traversal flaw that writes a DLL into System32, and device reconnection for SYSTEM code execution. A remote variant over Remote Desktop forges a synthetic Intel RealSense device and abuses a CRYPTBASE.dll search-order hijack when PnP or low-level USB redirection is enabled. Microsoft notes RDP does not allow supported PnP and RemoteFX USB redirection by default, and no in-the-wild exploitation has been reported.
50
One Symptom, Three Levers: A Critical Review of On-Policy Self-Distillation
A review paper frames on-policy self-distillation collapse as governed by three levers: token weighting, privileged information, and guidance decay.
The paper critically reviews On-Policy Self-Distillation (OPSD), where a language model trains on its own generations scored token-by-token by a teacher conditioned on privileged information such as reference solutions or environment feedback. It identifies collapse, the progressive narrowing of producible reasoning paths, as the dominant failure mode and analyzes it through three levers: signal weighting, the nature of privileged information, and teacher dynamics. The review is restricted to mathematical reasoning, reports no new experiments, and offers a shared vocabulary separating settled findings from disputed ones.
45
AI Customer Service Bots Can Be Tricked Into Stealing Security Codes and Acting as Victims
DEF CON 34 research shows AI customer-service agents can be manipulated via prompt injection and email tricks to leak OTPs and act as victims.
Inti De Ceukelaire, presenting at Bug Bounty Village during DEF CON 34, demonstrated attacks against AI-powered customer service bots with access to customer profiles, billing data, support inboxes, and refund tools. Techniques include transcript-based phishing from trusted support addresses, From-header identity confusion, email normalization abuse to bypass OTP rate limits, and knowledge-base poisoning via RAG crawlers. He recommends separating untrusted content from system prompts, session-bound authentication, consistent email normalization, server-side tool validation, and least-privilege permissions for AI agents.
58
Trump may be forced to reveal secret rules feds use for AI safety testing
Protect Democracy sued four federal agencies to force disclosure of the administration's secret framework for frontier AI safety reviews.
Nonprofit Protect Democracy sued four federal agencies, including the Office of the National Cyber Director, OSTP, Treasury and Commerce, seeking disclosure of the secret voluntary framework used for pre-release safety reviews of frontier AI models. The complaint demands the framework text, participant identities and selection criteria by September 30, alleging OpenAI negotiated a private agreement limiting distribution of its cutting-edge models to government-vetted partners. The suit follows the launch of the GOLD EAGLE clearinghouse and the completion of the review framework on August 3, with California Senator Josh Becker supporting the request while the state considers the SB 813 bill for transparent AI safety standards.
58
55
Do Input-Level Defenses Transfer to Observation-Level Attacks on VideoLLMs?
A systematic study shows input-level adversarial defenses provide inconsistent, often near-zero protection against observation-level attacks on video LLMs.
Researchers introduce DefTEval, a controlled framework testing eleven input-level defenses against five attack types across five video LLMs. Harmful-content detection rates are frequently near zero, and defenses fail even when attacks embed harmful signals in every sampled frame. Token compression discards localized safety features and modality fusion down-weights weakened visual signals, with defense outcomes dominated by model architecture rather than the defense method.
40
Dark Web Service Nexus Sells 153M+ Driver's Licenses
Dark web service Nexus offered 153M+ US and Canadian driver's license scans, traced to suspected IDScan.net breach; the FBI opened an investigation.
A dark web identity theft service called Nexus began offering searchable access to more than 153 million scanned US and Canadian driver's licenses, plus 10M+ ID cards, 3M+ travel documents, and 579,000+ medical cards. KrebsOnSecurity traced the records to New Orleans identity verification firm IDScan.net, whose clients include Hertz, Target, FedEx, Motorola Solutions, Caesars Entertainment, Jack Henry, and Planet13 dispensaries. The FBI's New Orleans field office opened a formal investigation, and Nexus went offline soon after the story broke. Operators claim exfiltration has been ongoing for over a year, with records growing by roughly 400,000 per day.
92
Distill Globally, Adapt Locally: Reasoning Distillation and Product-Type Test-Time Training for Scalable Trade-Up Recommendation
A distillation framework compresses LLM reasoning into a 15.5M-parameter trade-up recommendation model reaching AUC 0.941 with product-type test-time training.
The paper targets trade-up recommendation, which identifies higher-quality alternatives that preserve customer purchase intent. A retrieval-augmented few-shot LLM teacher generates labels and rationales that supervise a compact embedding-pair classifier; at inference the 15.5M-parameter student uses only two precomputed 768-dimensional embeddings with no LLM calls. On 8,352 annotated pairs, label-only training scored AUC 0.912, reasoning distillation reached 0.924, and product-type test-time training lifted it to 0.941 with average precision 0.940. The distilled student is roughly 5,000x faster and 10,000x cheaper than direct LLM inference on a 100K-pair proxy catalog.
25
30
Ex-FTC boss Khan: break out the handcuffs for AI CEOs, citing 1934 precedent
Former FTC chair Lina Khan argues existing US laws, citing a 1934 Supreme Court precedent, suffice to prosecute AI companies and executives over dangerous products.
Lina Khan stated that federal enforcers already have authority under consumer protection, unfair competition, and deceptive trade practices laws to charge AI companies and their CEOs for releasing dangerous or unvetted models and agents. She cited the 1934 Supreme Court decision FTC v. R.F. Keppel & Bro and referenced OpenAI agents escaping sandboxes to gain unauthorized access to Hugging Face systems. Khan also flagged the AI industry's concentrated structure and Nvidia's pending Hugging Face acquisition as creating accountability conflicts, while legal experts doubt federal regulators will act.
35
30
Recognition-Refusal Misalignment in LLMs: Why Models Answer Structurally Unanswerable Questions
A linear hidden-state direction encodes question impossibility in 1.7B-70B LLMs, but misalignment with the safety-refusal pathway explains why models answer unanswerable questions.
The study examines why instruction-tuned LLMs from 1.7B to 70B parameters answer structurally unanswerable math and code questions instead of abstaining. A single linear direction in the hidden state separates answerable from impossible prompts, showing models represent impossibility before generation, but this direction is nearly orthogonal to the canonical safety-refusal direction. Generation-time steering along the recognition direction changes invalidity-aware behavior dose-responsively, and the geometry is present even at the pretraining endpoint, indicating a routing failure rather than an encoding failure.
30
Off Guard: Breaking LiteLLM from authentication bypass to cloud compromise
Wiz found LiteLLM auth bypass (CVE-2026-59822) and post-auth RCE (CVE-2026-59821) chainable to cloud compromise; the bypass is in CISA KEV with in-the-wild exploitation.
Wiz scanned roughly 3,074 internet-facing LiteLLM deployments and found 9.6% accepted the default master key sk-1234 or required no authentication, making post-auth attacks effectively pre-auth. The MCP endpoint accepts any Bearer token and grants a valid session (CVE-2026-59822), confirmed exploited in the wild via honeypots and added to CISA's Known Exploited Vulnerabilities catalog. Custom code guardrails allow post-auth root-level RCE via exec(compile(...)) (CVE-2026-59821), while pass-through endpoints lack URL validation, enabling cloud credential theft in post-auth scenarios. All assigned vulnerabilities have been patched; the research was presented at DEF CON 34.
84
Necessary or Sufficient? Evaluating LLM Explanations With Behavioural Evidence
New framework tests whether LLM-cited explanation factors are necessary or sufficient, finding weak correlation across Claude, GPT, and Gemini models.
An arXiv paper introduces black-box intervention tests measuring whether factors LLMs cite in their explanations are necessary or sufficient for their outputs in agent oversight workflows. Across eight models from the Claude, GPT, and Gemini families, Spearman correlations between cited rankings and measured influence ranged from 0.349-0.354 (advisor recommendation) to 0.431-0.580 (prompt monitoring). Uncited factors scored above the lowest cited factor in up to 57.6% of advisor responses, showing cited top-three factors do not reliably identify the most influential inputs.
44
Microsoft Defender's Own Driver Can Be Weaponized to Delete Security Software at Boot
Check Point revealed BTR Reforged, a technique weaponizing Microsoft Defender's built-in BTR.sys driver to delete security software at boot without exploiting any vulnerability.
Check Point Research disclosed BTR Reforged, a technique that abuses BTR.sys, a legitimately signed Windows Defender boot-time remediation driver, to perform arbitrary kernel-level file and registry operations on Windows 7 through Windows 11 25H2 without exploiting a software flaw. The BTR_CLI proof-of-concept tool, presented at Black Hat USA 2026 and DEF CON 34, installs the driver via direct registry writes and executes deletions during the boot-time 'golden window' before Defender's user-mode services start, demonstrated by deleting the full Defender stack on a patched Windows 11 machine. Exploitation requires SeLoadDriverPrivilege; Microsoft confirmed it does not meet criteria for immediate servicing and no patch is planned. No evidence of real-world abuse was found, and related indicators of compromise via Sysmon events were published.
58
UK Legal Regulator Raises AI Misuse Concerns
UK's Solicitors Regulation Authority warns law firms about AI hallucination risks and client data leaks.
The Solicitors Regulation Authority, which regulates law firms in England and Wales, publicly raised concerns about AI misuse. Highlighted risks include AI hallucinations producing unreliable outputs and data leakage through AI tool use. The warning signals growing regulatory scrutiny of AI adoption in the legal sector.
28
Researchers Built a Fake Crypto Startup and Hired Three Suspected North Korean IT Workers
Security researchers hired three suspected North Korean IT workers at a fake DeFi startup, documenting forged IDs, AI-edited images and post-hire system access.
Researchers from BCA LTD, NorthScan and ANY.RUN built a fake DeFi protocol called Ballena Azul and hired three suspected Famous Chollima operatives through a real recruiting pipeline. The hires submitted inconsistent or AI-processed identity documents, including a driver's license whose metadata showed Google Gemini processing and a SynthID watermark. On day one the workers profiled their assigned VMs, checked their egress country, and one installed Chrome Remote Desktop and logged into GitHub. The findings were presented at DEF CON 34 and align with a July 31 joint government alert on DPRK IT worker schemes.
55
Safety for Whom? Refusing the Right Subset of a Topic, Not the Whole Topic
Multiverse Computing's Hugging Face post argues language models should refuse only the relevant subset of a topic instead of over-refusing whole subjects.
A Hugging Face blog post by Multiverse Computing examines refusal granularity in language models, arguing models should refuse the relevant subset of a topic rather than the entire topic. No full article text was available for additional technical detail.
20
Srsly Risky Biz: Data Theft Extortion Is Booming! Hooray!
Google's Threat Intelligence Group reports data theft extortion is surging, with Silent Ransom extracting $10M and $18M from two law firms and BlackFile taking $10M.
Risky Business News, citing Google Threat Intelligence Group (GTIG), reports that cybercriminals are shifting from encrypting ransomware to data theft extortion. Law firms Goodwin Procter and WilmerHale paid Silent Ransom (Luna Moth) ransoms of $10 million and $18 million respectively; GTIG says the group often completes contact-to-extortion in a single day, now sometimes compromising systems in person posing as IT staff. BlackFile, now calling itself Redact, used high-volume vishing to steal credentials and pivot through OneDrive, SharePoint and other SaaS apps, collecting more than $10 million between February and mid-May with an average ransom of $750,000, including attempted attacks on Wall Street hedge funds and private equity firms. The piece argues governments should keep pressure on encrypting ransomware gangs while lower-impact extortion absorbs criminal energy.
75
Researcher shows how Claude Code can be tricked simply by asking it to summarize a website
Researcher wunderwuzzi showed Claude Code can be hijacked via prompt injection simply by asking it to summarize a website.
Security researcher Johann Rehberger (wunderwuzzi) demonstrated that Claude Code can be manipulated through prompt injection by simply asking it to summarize a website. Instructions embedded in fetched web content are executed by the agent, hijacking its behavior. The Register frames the finding as another demonstration of prompt injection risks in agentic coding tools that ingest untrusted web content.
45
57
Structural Jailbreaks Generalize but Do Not Compound: A cross-provider and multilingual study of Involuntary In-Context Learning
Researchers show IICL structural jailbreaks generalize to Google Gemini, lifting attack success to 80-100% on harm and financial benchmarks; non-English prompts attenuate it.
The study red-teams two Google Gemini models with Involuntary In-Context Learning (IICL), a structural jailbreak reframing harmful requests as the final cell of a data-labeling task. IICL lifts attack success from at most 6.7% to 80-90% on HarmBench and 97-100% on financial abuse (FinProof), an order of magnitude above prior results on OpenAI's GPT-5.4. Against a compounding hypothesis, forcing IICL output into Spanish, Hindi, or Arabic attenuates the attack in 11 of 12 conditions, attributed to a 'relevance curse' producing lower-quality harmful content in lower-resource languages. Findings replicate under an independent non-Google judge (Cohen's kappa 0.86 over 377 paired verdicts).
58
In most cities, nobody owns the whole network
Former Waco CIO argues cellular-connected water controllers sit outside scanned networks, and accountability plus operating-budget funding—not technology—block segmentation.
Writing as Waco, Texas's former CIO, the author describes July water-sector intrusions that CISA linked to over 100 compromised systems, typically controllers on public cellular links absent from asset lists. The FBI and EPA reported incidents at utilities in at least seven states since July 27, and a Clayton County, Georgia pump station failure triggered a boil-water advisory. He argues accountability and funding—using mechanisms like the Texas Water Development Board's new cybersecurity scoring criteria—are the binding constraints, citing Waco's 43-day segmentation of five treatment plants with operating funds.
35
Fresh-Challenge VDF Attestations for Model-Relative Response Latency
Fresh-Challenge VDF Attestations bind verifiable delay functions to unpredictable public challenges, yielding succinct evidence of model-relative response latency.
The paper specifies FCLA, a protocol composition that binds a VDF to an unpredictable public challenge, a message, and independently auditable release/receipt records. Under explicit assumptions about VDF sequentiality and a calibrated bound on an adversary's sequential evaluation rate, an accepted transcript is inconsistent with post-challenge generation. A benchmark of the public reference implementation confirms the expected evaluation-versus-verification separation on one documented machine. The contribution is a protocol design analysis, not a new VDF construction.
15