Search: “Response Options”

Pick Your Poison: Learning to Select Poison Sets for Stronger LLM Backdoor Attacks

SAILS learns to select poison sets for LLM backdoor attacks, showing attack success ranges 3% to 80% at fixed poison counts across LLaMA-3-8B settings.

The paper shows existing backdoor evaluations that randomly sample a fixed number of poisoned examples severely underestimate worst-case vulnerability: across three LLaMA-3-8B settings, attack success ranges from 3% to 80% depending only on which poison set is chosen. SAILS formalizes poison selection as oracle-budgeted set optimization, learning a set scorer from a few hundred finetune-and-evaluate runs to rank millions of candidate sets and audit a small shortlist. It improves held-out attack success by 30 percentage points over the strongest influence baselines and transfers from small-scale to full-scale finetuning, extending to code-generation, agentic, and API-only backdoors.

arXiv cs.CR·Sep 14, 04:44 UTC · 2d agoAI safety & security 2 sources

The 12 Best Extended Detection & Response (XDR) Platforms, Compared and Priced

Buyer's guide compares 12 XDR platforms, favoring Microsoft Defender XDR, Stellar Cyber and CrowdStrike, and warns ingestion pricing inflates costs.

The article compares 12 extended detection and response platforms, distinguishing native XDR (CrowdStrike, Palo Alto Cortex XDR, Microsoft Defender XDR, SentinelOne) from open XDR (Stellar Cyber, Arctic Wolf, Rapid7). It argues data ingestion pricing, not per-endpoint fees, is the main budget risk and should be modeled before signing. It repeats the consolidation note that Sophos acquired Secureworks for approximately $859 million in February 2025, and flags that ExtraHop is NDR rather than full XDR.

GBHackers·updated Sep 9, 08:07 UTC · 7d ago·first Sep 9, 07:51 UTC · 7d agoIndustry 3 sources1

Building a ransomware decision tree before the call comes in

Arctic Wolf's IR VP urges pre-deciding ransomware response choices on containment, negotiation authority, law enforcement engagement, and crisis communications.

A Help Net Security video features Kerri Shafer-Page, VP of Incident Response at Arctic Wolf, walking through a ransomware decision tree. She recommends settling four decision areas in advance: containment, extortion negotiation limits, law enforcement involvement, and communications. She frames paying as a business decision that can cost less than insurance retention and renewal hikes, and notes agencies may know the threat actor and help avoid sanctions issues.

Help Net Security·Sep 11, 05:30 UTC · 5d agoIndustry

CREST Onboards First Cohort for AI-Enabled Pentesting Accreditation

CREST accredited 10 companies, including ImmuniWeb and Thoropass, in the first cohort of its new AI-Enabled Penetration Testing accreditation.

CREST granted its AI-Enabled Penetration Testing accreditation, an optional module added in July 2026 to its Penetration Testing Accreditation Standard, to a first cohort of 10 firms spanning Europe, India, the US, and Canada, including Closed Door Security, ImmuniWeb, JUMPSEC, Packetlabs, Pentesys, REDSECLABS, Risk Associates, SECNORA, Solusec, and Thoropass. The module lets providers that integrate AI into pentesting demonstrate responsible AI governance through independent assessment. A March 2026 CREST report found 76% of cybersecurity providers increased AI usage year over year and 69% already integrate it into daily service delivery; its June AI Charter has been signed by over 100 organizations.

Infosecurity Magazine·Sep 3, 13:00 UTC · 13d agoIndustry

The Frontier AEO Tracker: What Astra Chooses (and every other frontier model, and what you can do about it)

Latent Space launches an AEO tracker scoring 7 frontier models' product recommendations across 161 categories, revealing generational bias flips.

Latent Space built a tracker measuring Answer Engine Optimization by running 6 prompt variations across 7 frontier models with search enabled over 161 product categories, scoring first choices, alternatives, mentions, and anti-recommendations. It found 28 categories with a universally dominant primary choice and observed soft biases, such as models favoring their own lab's coding agents. Analysis of Anthropic's Sol→Astra and Opus→Fable generations showed newer models consulting fewer sources and being less likely to change answers when questions are paraphrased.

Latent Space·Sep 7, 21:32 UTC · 8d agoAI research

Top 10 Best Cloud Detection & Response (CDR) Solutions in 2026

Editorial scorecard ranks ten 2026 cloud detection and response platforms; Sysdig, Wiz, and CrowdStrike lead, with Wiz's Gem Security acquisition highlighted.

The editorial scorecard rates ten CDR platforms on real-time detection (30%), cloud telemetry depth, response automation, correlation, and value. Sysdig earns the best real-time detection score for its Falco- and eBPF-powered runtime telemetry, Wiz (8.7) folds acquired Gem Security's real-time CDR into its security graph, and CrowdStrike (8.7) leads response automation. Specialists Stream.Security, Skyhawk Security, Sweet Security, and the open-source Falco project are also assessed.

Cyber Security News·07:41 UTC · 7h agoIndustry

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.

arXiv cs.AI / cs.LG / cs.CL·Sep 15, 17:52 UTC · 21h agoAI research

Android car head units infected with proxy botnet malware through built-in software updaters

Kaspersky found malware delivered via car head unit updaters, attributed to the MoYu Group's BADBOX operation, recruiting devices into a proxy botnet.

Kaspersky discovered malware delivered through the built-in TWCore system updater in Android-based car head units running DoFun infotainment firmware, turning devices into ad-fraud tools and nodes in a proxy botnet. The three-stage infection chain (JarService dropper, loader, and final payload supporting nine commands) installs the zhima reverse-proxy module, which Nokia's Deepfield team independently found on TV set-top boxes. Kaspersky attributes the operation with high confidence to the MoYu Group, linked to the BADBOX supply-chain botnet first identified by HUMAN Security in 2023. DoFun closed the gap after Kaspersky's responsible disclosure.

Help Net Security·Aug 24, 00:00 UTC · 23d agoMalware in the wild

From Parameters to Answers: How LLMs Retrieve and Use Their Internal Knowledge

Interpretability study traces how Qwen, Llama, and Gemma route query information and internal knowledge across layers when answering questions.

Researchers used layerwise interventions on hidden states to separate query-routing signals from target knowledge in language models, testing Qwen, Llama, and Gemma on country-continent questions with varied answer types. A pair-conditioned request direction strengthens before interventions alter downstream knowledge, opening a causal window while answer-supporting content is still forming. Trajectories differ by model: Gemma shows a partially overlapping mid-layer routing profile, while Llama has no sustained routing-effect window under the same gates.

arXiv cs.AI / cs.LG / cs.CL·Sep 10, 17:39 UTC · 5d agoAI research1

EU's Cyber Resilience Act starts the 24-hour vulnerability clock

EU Cyber Resilience Act reporting rules take effect, requiring manufacturers to disclose actively exploited vulnerabilities to ENISA within 24 hours, with fines reaching €15 million.

The Cyber Resilience Act's Article 14 mandatory reporting duties became applicable, requiring makers of products with digital elements sold in the EU — regardless of where they are based — to file an early warning within 24 hours of becoming aware of an actively exploited vulnerability, a detailed notification within 72 hours, and a final report within 14 days of releasing a fix. Reports must be submitted through ENISA's Single Reporting Platform to the designated CSIRT, and non-compliance with these core duties can trigger fines up to €15 million or 2.5 percent of annual turnover. Manufacturers must also inform affected users of available fixes without undue delay, and most remaining CRA provisions, including mandatory SBOMs and security-by-design requirements, become applicable on December 11, 2027.

The Register · Security·Sep 11, 11:34 UTC · 5d agoPolicy & legal

Ask Before You Optimize: Dynamic Pre-Formulation Clarification for Interactive Optimization

Researchers release OR-Clarify, a benchmark testing whether LLM agents ask clarifying questions before formulating optimization models from incomplete requests.

OR-Clarify evaluates pre-formulation clarification in operations research: each task gives a partial problem description, withholds structured hidden slots, and scores agents via bounded interaction with a simulated user, measuring slot recovery, stopping behavior, silent assumptions, and interaction cost. The authors also propose InterOPT, a two-stage framework that identifies formulation-critical gaps to decide when to ask or stop. In choice-based experiments InterOPT substantially outperforms all baselines in exact slot recovery and remains competitive in the open-ended setting.

Hugging Face daily papers·Sep 3, 20:00 UTC · 12d agoAI research1

Your Cloud Security Checklist Doesn't Work the Way You Think It Does

Intruder's 2026 Cloud Security Index found misconfiguration risk profiles differ sharply across AWS, Azure, and Google Cloud across 3,000 organizations.

Intruder analyzed misconfiguration data from 3,000 organizations across AWS, Azure, and Google Cloud for its 2026 Cloud Security Index. Weak IAM controls and missing logging affected 80-98% of accounts regardless of provider, while exposed services ranged from 76% on AWS to just 8% on Google Cloud. Top issues included S3 buckets without HTTPS enforcement (87% of AWS accounts), Entra ID users without MFA (55% of Azure accounts), and missing OS Login MFA (77% of Google Cloud accounts). Weak IAM prevalence rose with organization size, from 87% at SMEs to 98% at large enterprises, and midmarket organizations took the longest to remediate at 35 days on average.

The Hacker News·Sep 7, 11:45 UTC · 9d agoResearch

llm 0.34

Version 0.34 of Simon Willison's llm CLI adds response-duration metrics to log output, plus bug fixes and faster log querying.

The open-source llm command-line tool for interacting with large language models released version 0.34. The headline change adds response duration in milliseconds and human-readable form to llm logs --usage Markdown output, plus a new duration_ms field in llm logs --short. The release includes several contributed bug fixes and a significant performance improvement to llm logs, alongside the related llm-openrouter 0.7.1 release.

Simon Willison·Sep 2, 19:23 UTC · 13d agoAI tools & infra1

Product showcase: Trust Chain TPRM turns vendor compliance evidence into verified assurance

Help Net Security·Aug 10, 12:30 UTC · Aug 10, 2026Policy & legal

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, 12:34 UTC · Aug 17, 2026AI safety & security

LG accused of 'egregious invasion of privacy' over TV data collection

Gamers Nexus alleges LG smart TVs record audio and generate transcripts in standby while harvesting location, network, and device data for LG's ads business.

Researchers at Gamers Nexus claim LG smart TVs continued capturing audio after voice recognition activated, including in standby, producing plaintext transcripts, some stored locally and sent after reconnection. Testing reportedly found collection of IP addresses, location data, nearby Wi-Fi network details, and enumeration of unpaired local devices such as phones, routers, and PCs, with data flowing to LG Ads Solutions. The team is coordinating responsible disclosure of vulnerabilities including an alleged remote code execution flaw. LG says its TVs do not collect, record, or store ambient conversations and that voice recognition is optional.

The Register · Security·Sep 8, 14:15 UTC · 8d agoIndustry1

Approval Integrity and Recovery in LLM Answer Publication

Study measures approval integrity in Lightcap LLM answer publication, finding the 14B response-act checker accepts 291 of 302 unsupported answers.

The study evaluates exact-content binding, authorization freshness, and checkpoint recovery in Lightcap's publication enforcement using 3,600 assessments over 900 human-annotated RAGTruth responses from three Ministral models. The production 14B response-act checker accepts 291 of 302 unsupported answers versus 41 for a direct-grounding baseline, with supported-answer retention of 95.2% versus 66.9%. A stateful recheck-recovery policy increases exact-match error by 9.23 percentage points relative to initial checkpoints, and controlled evidence-fingerprint changes expose asymmetric freshness enforcement between publication and recovery. A separate BIPIA prompt-injection experiment records zero target insertions among 266 valid editor outputs.

arXiv cs.CR·Sep 14, 13:48 UTC · 2d agoAI safety & security

AI is getting better at election facts, but voters shouldn’t rely on it

CyberScoop·Aug 5, 09:00 UTC · Aug 5, 2026Exploit / PoC in the wild

Rosetta at AlexandriaX-2026: LoRA-Adapted NileChat for Context-Aware Dialectal Arabic Dialogue Translation

Rosetta ranks 4th and 5th in AlexandriaX-2026 dialectal Arabic dialogue translation using a LoRA adapter on NileChat-3B, finding limited pretraining benefit.

The Rosetta system for the AlexandriaX-2026 shared task fine-tunes a LoRA adapter on NileChat-3B for context-aware English-to-dialectal Arabic dialogue translation. The adapter was additionally pretrained on MADAR and PADIC dialect corpora for the unconstrained track. It achieved spBLEU 26.10 (4th, constrained) and 25.09 (5th, unconstrained). External dialect pretraining improved only two of thirteen dialects while slightly degrading overall performance, indicating negative transfer.

arXiv cs.AI / cs.LG / cs.CL·Sep 9, 16:15 UTC · 6d agoAI research1

When AI Agents Go Rogue: Agent Session Smuggling Attack in A2A Systems

Unit 42 unveils agent session smuggling, where a rogue AI agent hides covert instructions in established Agent2Agent (A2A) protocol sessions to manipulate victim agents.

Palo Alto Networks Unit 42 discovered agent session smuggling, a new attack technique in which a malicious AI agent exploits an established cross-agent session under the Agent2Agent (A2A) protocol to send covert instructions hidden among benign client requests and server responses. The technique leverages the implicit trust agents place in collaborating agents and the stateful, multi-turn nature of A2A sessions; the researchers stress it affects any stateful protocol, not an A2A flaw. Unlike one-shot data-based attacks, a rogue agent can converse, adapt and build false trust over multiple interactions. Proposed mitigations include human-in-the-loop enforcement, cryptographically signed AgentCards for remote agent verification, and context-grounding to detect injected instructions.

Palo Alto Unit 42·Aug 17, 12:32 UTC · Aug 17, 2026AI safety & security2

When the prompt becomes the payload: A practical pen-testing guide for GenAI, LLM and RAG applications

CSO Online publishes a practical penetration-testing guide for GenAI, LLM, and RAG applications, covering prompt injection, retrieval poisoning, and tenant isolation testing.

The guide frames LLM applications as attack graphs spanning prompts, retrieval layers, vector stores, tools, identities, and downstream APIs, arguing that conventional web testing misses instruction-vs-data channel risks. It builds on OWASP prompt injection guidance (direct vs. indirect injection) and NIST's 2025 adversarial machine-learning taxonomy, noting that RAG and fine-tuning do not remove injection risk. Recommended practices include documenting trust transitions across components, using canaries and synthetic records to avoid test side effects, running multi-turn and obfuscated injection campaigns, and verifying chains from poisoned documents to observable state changes. It also details testing RAG pipelines via controlled document poisoning across metadata, OCR layers, and code comments, plus cross-tenant isolation checks on retrieved document IDs.

CSO Online·Sep 9, 10:00 UTC · 7d agoAI safety & security1

iAuthFlow v2: The $10,000 Phishing Toolkit That Survives Your Password Reset

Abnormal Security details iAuthFlow v2, a $10,000 phishing kit that enrolls attacker-controlled passkeys on Google accounts, persisting through password resets.

Abnormal Security analyzed iAuthFlow v2, a phishing toolkit sold on a Russian-language cybercrime forum for a $10,000 base price with additional capability modules sold separately. Using a browser-in-the-middle relay behind a trycloudflare.com subdomain with valid TLS, it captures Google logins and uses the authenticated session to enroll an attacker-controlled passkey within six seconds of login. Because the passkey persists after password changes and session revocation, operators regain mailbox access even after victims reset credentials. The build examined targets Google, while the seller advertises versions for Microsoft, iCloud, and LinkedIn.

Security Affairs·Aug 24, 07:17 UTC · 23d agoPhishing & fraud1

Operational Resilience: IT Security Risks with Reduced Staffing | Huntress

Huntress blog advises security teams on managing change, risk, and incident response during reduced-staffing holiday periods.

The article discusses how holiday-period staffing reductions change organizational risk profiles around change management, monitoring, and incident response capability. It argues against blanket change freezes when critical vulnerabilities with high exploitation probability demand patching, and stresses retaining decision-making authority, escalation paths, and recovery knowledge. It concludes by promoting Huntress Managed Response, which lets the Huntress SOC take predefined containment actions on confirmed threats without customer intervention.

Huntress·Sep 15, 14:00 UTC · 1d agoIndustry

Hugging Face Diffusers Flaws Could Let Model Repositories Execute Arbitrary Code

The Hacker News·Aug 3, 06:40 UTC · Aug 3, 2026VulnerabilityCVE-2026-44827CVE-2026-45804CVE-2026-445131

PACE: Perceived-Latency-Aware Cascading Service Routing and Filler Control for QoE-Efficient Retrieval-Augmented Dialogue Serving

PACE cuts perceived latency in retrieval-augmented dialogue serving via cascading routing and filler control, reaching 0.41s P95 under load.

PACE is a serving framework for retrieval-augmented dialogue that optimizes Perceived Time-to-First-Response (PTFR) under quality and cost constraints. It combines a load-adaptive cascading router, a joint path-filler controller, and volatility-aware cache admission, deployed on a humanoid-robot sales service. On 75k CarQA requests, the cascade halved pure-LLM P95 PTFR (0.29s vs 0.53s) and the adaptive controller reached 0.41s P95, 2.4x better than RAG at high load; filler calls dropped 94% and stale answers fell from 86% to 0%.

arXiv cs.AI / cs.LG / cs.CL·Sep 9, 16:02 UTC · 6d agoAI tools & infra

Debian developers rejected an LLM ban and left disclosure voluntary

Debian developers voted to encourage voluntary disclosure of AI assistance in contributions rather than banning or mandating labeling of LLM-generated code.

Debian's vote concluded August 28, with project secretary Kurt Roeckx announcing that the winning option encourages contributors to disclose generative AI assistance and stops there, adding no new mandatory review gates. The adopted resolution bars sending confidential material, embargoed security bugs, cryptographic keys and credentials to third-party AI services without explicit authorization, requires prior discussion for bulk automated work, and takes no position on whether model output is copyrightable. The project neither endorses nor prohibits generative AI and retains existing human review and licensing requirements.

Help Net Security·Aug 31, 00:00 UTC · 16d agoAI policy1

You Can't Prefer Emotions You Don't Sample: Intensity Undershoot in DPO-Tuned LLMs

Study quantifies DPO-tuned LLMs undershooting requested emotional intensity, tracing the gap to candidate-pool extremity rather than conditioning format.

Conditioning an instruction-tuned LLM on continuous valence-arousal targets yields gain of only 0.26 for valence and 0.13 for arousal on Llama-3.1-8B, far below faithful control of 1.0. The authors attribute undershoot to neutral-heavy preference corpora like EmoBank and candidate pools lacking extreme affect, leaving DPO without extreme exemplars. Uniform target coverage with a hotter candidate pool raises valence gain to 0.40 on Llama-3.1-8B and 0.44 on Qwen3-8B, with modest in-distribution cost; arousal gains remain unstable across seeds.

arXiv cs.AI / cs.LG / cs.CL·Sep 7, 17:47 UTC · 8d agoAI research

Cisco Fixed Critical RCE in Nexus 9000 Series Switches

Cisco patched CVE-2026-20212 (CVSS 9.8) in Silicon One-based Nexus 9000 switches, allowing unauthenticated remote root code execution via TCP ports 43210/43211.

Cisco patched CVE-2026-20212 (CVSS 9.8), a flaw in the Silicon One integration for Nexus 9000 Series switches that lets unauthenticated remote attackers execute code with root privileges. TCP ports 43210 and 43211 are exposed through the default Layer 3 VRF, and exploitation can also crash the S1HAL process, forcing device reloads. Cisco TAC discovered the flaw during a support case; PSIRT is not aware of public disclosure or malicious exploitation. Workarounds include infrastructure ACLs or blocking the exposed ports, alongside a Live Protect shield pending fixed NX-OS upgrades.

Security Affairs·Sep 3, 19:49 UTC · 12d agoVulnerabilityCVE-2026-20212

Trends in Web Threats in CY Q2 2022: Malicious JavaScript Downloaders Are Evolving

Unit 42 detected 751,000 landing URL incidents in Q2 2022 and documented malicious JavaScript downloaders evolving to evade detection.

Unit 42 detected 751,331 landing URL incidents (253,644 unique) and 1,744,629 malicious host URL incidents (256,844 unique) from April through June 2022. Total landing URL incidents rose compared with Q1 2022, and unique host URL incidents grew 42%, indicating attackers deploying more variants. The report includes a case study of a JavaScript downloader campaign demonstrating new evasion techniques. Personal sites, blogs, and business sites were the top apparently benign entry points.

Palo Alto Unit 42·Aug 17, 13:05 UTC · Aug 17, 2026Research

Exposed Vite servers are being probed for AWS and Azure credentials

F5 honeypots logged 32,000+ probes against Vite file-access bypass CVE-2026-39364, hunting AWS and Azure credentials on exposed dev servers.

F5 Labs reported 32,000+ scan attempts (807 attack sessions) against exposed Vite servers in August, up from 1,732 attempts over the prior three months. Attackers exploited CVE-2026-39364 (CVSS 8.2), which bypasses Vite's server.fs.deny protection via parameters like ?raw and ?import&raw, affecting Vite 7.1.0 to before 7.3.2 and Vite 8 before 8.0.5. Scanners cycled wordlists for environment files, AWS keys, Azure tokens and IaC state files, also combining older Vite CVEs (2025-30208, 2025-31125, 2024-45811) and probing a Next.js middleware bypass. Only CVE-2025-31125 is currently in CISA's KEV catalog.

CSO Online·Sep 15, 14:36 UTC · 1d agoExploit / PoC in the wildCVE-2026-39364CVE-2025-30208CVE-2025-31125+4 CVEs

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·Sep 7, 20:00 UTC · 8d agoAI research

Expanding AI access and cyber defense for federal, state, local, and tribal governments

OpenAI and GSA agreed to provide ChatGPT to federal, state, local, and tribal governments with $0 licenses, 50% off usage, and cyber-defense access.

OpenAI for Government and the U.S. General Services Administration announced a 27-month agreement (October 1, 2026 through December 31, 2028) waiving the $15 per-user monthly license fee and cutting usage costs 50% for federal, state, local, and tribal agencies. More than one million government employees already have ChatGPT access, with eligibility extending across a roughly 23 million-person U.S. public-sector workforce. Every verified government entity is approved for Daybreak Blue cyber-defender access at 50% off, with Daybreak Red available for vulnerability research, exploit validation, and red teaming at standard pricing. The deal builds on the $1 billion Daybreak for Frontline Defenders commitment announced the prior week.

OpenAI News·updated Sep 10, 18:25 UTC · 5d ago·first Sep 10, 07:00 UTC · 6d agoAI industry 3 sources1

A Deep Dive Into Attempted Exploitation of CVE-2023

Mirai-like botnet scans exploit TP-Link EOL router flaw CVE-2023-33538 after CISA KEV addition, though observed exploit code is flawed.

Unit 42 observed large-scale automated scans attempting to exploit CVE-2023-33538 in end-of-life TP-Link TL-WR940N, TL-WR740N and TL-WR841N routers after CISA added the flaw to its KEV catalog in June 2025. HTTP GET requests inject commands via the ssid1 parameter at the /userRpm/WlanNetworkRpm endpoint to download and execute an arm7 ELF binary, a Mirai variant related to the Condi IoT botnet. Firmware emulation and reverse engineering showed the observed exploits are flawed and would fail, but the underlying vulnerability is real and successful exploitation requires authentication to the router's web interface. TP-Link confirmed the devices are end-of-life with no patches available and recommends replacing units and eliminating default credentials.

Palo Alto Unit 42·Aug 19, 12:03 UTC · 28d agoExploit / PoC in the wildCVE-2023-335381

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.

Hugging Face daily papers·Sep 2, 20:00 UTC · 13d agoAI safety & security1

Credentialed Pre-Port Discovery: Don't Probe the Host, Ask it

Rapid7's version 8.58 adds credentialed pre-port discovery, letting scan engines query hosts for listening TCP/UDP ports instead of network probing.

Rapid7 documents credentialed pre-port discovery, available from version 8.58 as an off-by-default per-template option under Asset Discovery. The scan engine authenticates with existing SSH, Scan Assistant, or Windows credentials and receives the host's authoritative list of listening TCP and UDP ports, skipping the network port scan. Hosts without matching credentials silently fall back to network port scanning. Trade-offs include reporting firewall-blocked ports the engine cannot reach, longer scans, and early termination findings caused by repeated connection failures.

Rapid7 Blog·Sep 9, 15:16 UTC · 6d agoTools

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.

Hugging Face daily papers·Aug 28, 20:00 UTC · 18d agoAI safety & security

ukisai/Swift-Qwen3.8-27b — new model trending #30 on Hugging Face

UkisAI releases Swift-Qwen3.8-27B, a Qwen3.8-27B derivative using 58.3% fewer thinking tokens with <1% performance loss and ~1.95x speed-up.

UkisAI released Swift-Qwen3.8-27B, a reasoning-efficient derivative of Qwen3.8-27B that cuts thinking-token usage by 58.3% while staying within 1% of base performance, yielding a 1.95x speed-up on several tasks. The model was fine-tuned by penalizing reasoning-marker tokens that trigger overthinking, plus a transfer component from BottleCap AI's ThinkingCap-Qwen3.6-27B. Benchmarks include GPQA-Diamond 88.28% (base 88.38%), MMLU-Pro 84.95% (base 85.47%), and AIME 2026 94.00% (base 98.67%), with mean-token reductions of roughly 27-46% across tests. GGUF weights are available on Hugging Face alongside enterprise licensing options.

Hugging Face trending models·Sep 8, 13:35 UTC · 8d agoModel release

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.

Hugging Face Blog·Sep 8, 14:23 UTC · 8d agoAI safety & security

Multiple Vulnerabilities Discovered in a SCADA System

Unit 42 details five vulnerabilities (CVSS 7.0-7.8) in ICONICS Suite SCADA software enabling privilege escalation and DoS.

Unit 42 discovered five vulnerabilities (CVE-2024-1182, CVE-2024-7587, CVE-2024-8299, CVE-2024-8300, CVE-2024-9852) in ICONICS Suite versions 10.97.2 and earlier for Windows during a 2024 security assessment. The flaws, rated CVSS 7.0-7.8, allow DLL hijacking, privilege escalation, information disclosure, denial-of-service and potentially full system compromise. ICONICS Suite is a SCADA solution with hundreds of thousands of installations in over 100 countries, widely used in critical infrastructure, and several dozen servers are internet-exposed per Unit 42 telemetry. ICONICS released patches and advisories with workarounds in 2024.

Palo Alto Unit 42·Aug 19, 12:04 UTC · 28d agoVulnerabilityCVE-2024-1182CVE-2024-7587CVE-2024-8299+2 CVEs1

What Did I Just Say? Self-Listening for Full-Duplex Speech Models

Researchers propose Self-Listening, a full-duplex speech approach feeding realized model speech back as input to improve interruption recovery.

Full-duplex spoken language models can listen and speak simultaneously, but asynchronous text generation, speech synthesis, and playback cause mismatches between what a model believes it said and what the user heard. The paper defines the resulting recovery problem as anchor interruption and proposes Self-Listening, which interleaves user speech, model text, and played speech as input streams. The authors also release AnchorSpeech, a benchmark with homogeneous training and test splits tracking which ordered response items were actually spoken. Experiments show self-listening models achieve better anchoring performance than full-duplex baselines.

Hugging Face daily papers·Sep 3, 20:00 UTC · 12d agoAI research