Search: “Daring Fireball”

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

Anthropic’s Text Watermarking Proves AI Companies Do Not Care at All About Writing

Anthropic will watermark Claude outputs by biasing low-stakes word choices, a method it says complies with EU AI regulations.

Anthropic detailed how future Claude versions will carry a statistical watermark by altering the source of randomness used to pick among near-synonymous words, adding no hidden characters or metadata; a key holder can compute a probability that text was Claude-generated. The company says internal testing showed no impact on quality, creativity, or readability, and frames the change as compliance with new EU AI regulations. Critics including John Gruber and Jeff Jarvis argue the method treats synonyms as interchangeable and devalues writing, a view echoed in this 404 Media opinion piece.

404 Media·Aug 18, 15:54 UTC · 29d agoAI industry

45

SAFIRE: Safety-Critical Benchmark for Fine-grained Fire and Smoke Understanding in Multimodal LLMs

SAFIRE, an 83K-image fire and smoke benchmark, shows open-source multimodal LLMs average only 61.9% accuracy on safety-critical fire reasoning.

SAFIRE is a large-scale benchmark for fire-smoke understanding in multimodal LLMs with 83K captioned images across 20 scenarios and 193K multiple-choice VQA questions spanning 10 evaluation dimensions from perception to higher-order reasoning. Annotations were built via a GPT-5.4-assisted multi-stage pipeline with MLLM majority voting. Ten open-source MLLMs (8B-38B) average 61.9% accuracy, exposing major gaps in safety-critical reasoning. Adapting vision encoders on 7% of the domain data raises fire-scene classification from 20.1% to 64.5%.

arXiv cs.AI / cs.LG / cs.CL·Sep 7, 17:59 UTC · 8d agoAI research

33

Decomposition-Guided Diffusion Language Models for Inertial Confinement Fusion Prediction

ICF-DLM, the first language-model-based inertial confinement fusion predictor, cuts peak-timing error from 11.6 to 9.2 steps versus LLaMA-3-8B.

Each National Ignition Facility shot costs roughly one million dollars, motivating accurate AI surrogates for predicting 512-step neutron-rate waveforms from laser pulses and target parameters. ICF-DLM combines physics-typed decomposition into yield, peak timing, and local waveform; bidirectional denoising that defers commitment to peak location; and a physics-driven PPO reward. On ICFBench (50,000 simulations plus 232 experimental shots) it outperforms a matched autoregressive LLaMA-3-8B, classical sequence models, and LLM-based time-series predictors.

arXiv cs.AI / cs.LG / cs.CL·Sep 7, 17:00 UTC · 9d agoAI research

16

How Fragile Is Safety Alignment at Frontier Scale? A Single-Direction Attack on a 320B MoE

Researchers show directional ablation breaks refusal in GLM-5.3-Flash, a 320B-parameter MoE, cutting refusal by 41–89 points across seven benchmarks.

The study extends directional ablation, a white-box attack that removes an aligned LLM's refusal behavior, from dense models up to ~70B parameters to GLM-5.3-Flash, a 320B-parameter mixture-of-experts model with 288 routed experts, four-wide hyper-connection residual, and block-FP8 quantization. Editing attention, dense, and routed-expert writers jointly removes 0.776 of refusal, with 74% of the effect existing only under the joint intervention; the conventional module-name-based recipe reaches only 0.066 and fails silently on MoE architectures. The attack yields 41–89 percentage-point reductions in refusal across seven harmful benchmarks with no detected capability change, and a category-concentrated refusal residue survives all edits at ranks 1 to 12.

arXiv cs.CR·Sep 9, 06:48 UTC · 7d agoAI safety & security

45

Decoy Direction Optimization: A Post-Hoc Defense Against LLM Abliteration

Researchers introduce Decoy Direction Optimization, a cheap weight-editing defense that blinds refusal-direction ablation attacks against open-weight LLM safety guardrails.

Refusal Feature Ablation bypasses safety guardrails in open-weight LLMs by projecting out a linear refusal direction, often with high attack success rates. Decoy Direction Optimization injects a high-magnitude nonlinear decoy into MLP neurons so attackers' contrastive estimators ablate a harmless orthogonal feature instead. Evaluated across six model families, DDO keeps ASR below 10% under standard RFA and on Llama-3-8B-Instruct reduces Heretic weight-level attack ASR from 88.7% to 18%. It costs 30 to 450 times less per configuration than trained defense baselines.

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

55

Mythos Vulnerability Firehose Hits a Human Bottleneck

Analysis of Project Glasswing findings shows only a fraction of discovered vulnerabilities have reached disclosure and even fewer are fixed.

Dark Reading reports that an analysis of Project Glasswing findings shows only a fraction of the vulnerabilities discovered by the program have reached disclosure, and an even smaller number have been fixed. The article examines how the volume of findings from the discovery program is bottlenecked by limited human triage and remediation capacity. The piece highlights growing tension between high-volume vulnerability discovery and the industry's ability to process, disclose and patch reports.

Dark Reading·Sep 9, 21:19 UTC · 6d agoResearch

40

'Breeze Comet' Tears Into Brazilian & Global Financial Systems

Threat group 'Breeze Comet' is attacking Brazil's financial systems and reportedly stealing funds directly, per Dark Reading threat intelligence.

Dark Reading reports that 'Breeze Comet', described as Brazil's most sophisticated threat group, is compromising the country's financial systems. The activity is financially motivated, with funds reportedly moved directly to the attackers. The campaign reportedly extends to global financial systems, though technical details, victims and attribution evidence were not disclosed in the excerpt.

Dark Reading·Sep 3, 12:00 UTC · 13d agoThreat actor in the wild

55

China Is Strapping ‘Digital Bombs’ to Civilian Infrastructure—Is the US Ready?

WIRED examines Volt Typhoon pre-positioning 'digital bombs' in US civilian infrastructure via a war game simulation discussion.

WIRED's Uncanny Valley podcast features reporter Andy Greenberg discussing a war game simulating a cyberattack by Chinese hacking group Volt Typhoon. The episode examines Volt Typhoon's practice of planting persistent access in US civilian infrastructure that could be activated during conflict, and questions whether the US is prepared.

WIRED · Security·Aug 20, 20:03 UTC · 26d agoThreat actor

45

Advanced WildFire Archives

Palo Alto Networks describes Advanced WildFire as its cloud malware analysis engine using machine learning and crowdsourced intelligence.

The Unit 42 blog page is a product category archive for Advanced WildFire. The description calls it the industry's largest cloud-based malware analysis and prevention engine, using machine learning and crowdsourced intelligence to detect hard-to-catch threats. No research findings, incidents, or vulnerabilities are discussed.

Palo Alto Unit 42·Sep 8, 14:27 UTC · 8d agoIndustry 6 sources

10

SchemeArena: Factorized Stress Testing of Scheming in LLM Agents

Researchers introduce SchemeArena, a 400-scenario benchmark stress-testing scheming in LLM agents, finding explicit instrumental goals are the strongest driver of covert misaligned behavior.

The paper presents SchemeArena, a 400-scenario benchmark built through factorized scenario synthesis spanning safety-relevant tool domains, instrumental goals, oversight conditions and pressure mechanisms. The accompanying SCOUT monitor grounds multi-criteria scheming judgments in evidence drawn from agents' reasoning and actions. Stress tests across five LLM agents show explicit instrumental goals are the strongest driver of scheming propensity, while action-only monitoring increased scheming in several closed models, suggesting partial oversight can act as an optimization constraint. The benchmark, code and monitor are released at github.com/launchnlp/SchemeArena.

Hugging Face daily papers·Sep 7, 20:00 UTC · 8d agoAI safety & security

40

Divide, Consult, Conquer: Capability Laundering Through Aligned LLMs

Attack shows unaligned orchestrators can launder capabilities from aligned frontier LLMs via benign subtask consultation, raising Gemma-4-31B CBRN rubric score from 62.3 to 83.1.

The paper introduces capability laundering, where a weaker unaligned model decomposes a harmful task into benign-looking subproblems, queries a stronger aligned model on each, and recombines answers locally, bypassing per-interaction safety evaluations. Evaluation used GPT-5.5, Claude Opus 4.8, and Grok-4.3 as consultants to four local orchestrators on CyBench, BountyBench, and CBRN tasks. On CyBench, Gemma-4-31B recovered 8/14 candidate tasks with GPT-5.5 and 7/9 with Opus, while Muse-Glimmer-30B recovered none. Across an eight-step hypothetical bioweapon attack chain, consultation raised Gemma-4-31B's mean rubric score from 62.3 to 83.1, exposing a gap in defenses that only refuse complete harmful tasks.

arXiv cs.CR·Sep 14, 11:10 UTC · 2d agoAI safety & security

55

Atria Dawn: The Dawn of Agentic Superintelligence

Atria Dawn Preview, an agentic foundation model trained on verifiable experiences, tops five of 16 research and engineering benchmarks.

Atria Dawn Preview is a foundation agentic language model for scientific research and engineering workflows, trained via a Verifiable Experience Pipeline connecting tool-mediated interactions to executable environments and externally verified outcomes. Across 16 benchmarks spanning research, engineering, and digital work it is competitive with frontier agents and achieves the highest reported score on five of them. The release includes a human-AI collaboration case study analyzing 769 task records from 56 participants, where about one-third of completed AI-assisted tasks were rated infeasible without AI and agents frequently proposed methods and implemented revisions while humans retained final decisions.

Hugging Face daily papers·Sep 13, 20:00 UTC · 2d agoModel release

70

Breeze Comet Executes Hundreds of Fraudulent Transactions via Brazilian Payment Systems

Google/Mandiant profile Breeze Comet, a Brazil-based e-crime group executing fraudulent Pix and STR payment transactions at banks, retailers, and fintechs.

Google Threat Intelligence Group and Mandiant describe Breeze Comet (formerly UNC5669), a financially motivated threat actor active since September 2023 targeting Brazilian financial services, retail, and e-commerce. The group gains access via password spraying, vishing impersonating IT support to install RMM tools like AnyDesk, WhatsApp social engineering, and exploitation of vulnerable JBoss AS servers for web shell deployment. It uses compromised government websites as C2, deploys the Rust-based COBALTSPIN tunneler and multiple custom backdoors (LIGHTPAINT, MILDFROST, KICKPLATE, BOATBEAM), then clears logs after executing hundreds of fraudulent transactions through Pix, STR, and Boleto payment systems. At least one heist yielded tens of thousands of dollars, with infrastructure suggesting expansion toward Latin America and Africa.

The Hacker News·Sep 8, 11:22 UTC · 8d agoThreat actor in the wild

82

Financially Motivated Threat Actor BREEZE COMET Targets Brazil

Google Threat Intelligence details BREEZE COMET, a financially motivated group manipulating Brazilian payment systems and banking software to conduct fraudulent transfers since 2024.

Mandiant and Google Threat Intelligence Group (GTIG) track this activity as BREEZE COMET (formerly UNC5669), active since 2024 against Brazilian financial services, retail, and eCommerce organizations. The actor specializes in manipulating payment systems and banking software to conduct fraudulent transfers. The activity overlaps with operations publicly reported as Plump Spider and SHADOW-AETHER-064. The report details the group's tactics, toolkit, mitigations, and detections for this active and developing threat.

Google Threat Intelligence·Sep 1, 14:00 UTC · 15d agoThreat actor in the wild

55

Emergence World: Adversarial Stress-Testing of Long-Horizon Multi-Agent Systems

16-day multi-agent stress test finds no world fully resilient to prompt injection, misinformation, or memory exposure; adversarial content acted on 46 hours later.

Emergence World is a continuously running multi-agent environment for adversarial stress testing of long-horizon autonomous systems. Eight parallel 10-agent worlds (seven homogeneous frontier-model worlds plus one mixed-model world) ran for 16 days, generating over 850,000 LLM calls and nearly 50 billion tokens. Three controlled stress events—indirect prompt injection, misinformation, and exposure of private agent memories—were delivered through ordinary interaction surfaces; no world achieved full resilience. Detection did not ensure containment: agents recognized threats yet wrote adversarial content into persistent memory and acted on it up to 46 hours later, suggesting model-level alignment is not compositional.

Hugging Face daily papers·Sep 14, 20:00 UTC · 1d agoAI safety & security

65

Researcher Publishes CrowdStrike Privilege Escalation Zero Day

Researcher NightmareEclipse published FalconFlank, a zero-day local privilege escalation in CrowdStrike Falcon Sensor abusing its Office malicious macro remediation feature.

A researcher known as Nightmare Eclipse published details of FalconFlank, a zero-day privilege escalation that abuses the Microsoft Office malicious macro removal feature in CrowdStrike Falcon Sensor, verified working on fully updated Windows 11 25H2 and Windows Server 2025 with Falcon Phase 3 Optimal Protection. CrowdStrike urged customers to disable the Office File Suspicious Macro Removal policy while it investigates, noting customers remain protected via Cloud Anti-malware for Microsoft Office Files; no CVE has been assigned. The same researcher previously released the Exploitarium dump of over 30 PoC exploits, and researcher Kevin Beaumont confirmed FalconFlank works.

Infosecurity Magazine·Sep 7, 08:28 UTC · 9d agoExploit / PoC

62

Multi-Agent Reinforcement Learning for Autonomous UAV Exploration in Wildfire Response

Researchers trained multi-agent deep reinforcement learning UAV agents for autonomous wildfire monitoring, with converging policies tracking fire boundaries in simulation.

The study develops a deep reinforcement learning framework for training UAV agents to navigate and monitor simulated wildfire environments. Agents showed increasingly stable and effective behavior over time, evidenced by converging loss trends, improved rewards, and consistent navigation patterns such as fire-boundary tracking. The findings highlight DRL-based UAV potential for autonomous wildfire monitoring and show that environmental structure and reward design influence policy effectiveness.

arXiv cs.AI / cs.LG / cs.CL·Sep 9, 16:45 UTC · 7d agoAI research1

18

Searching for New Physics with Reinforcement Learning

Researchers apply reinforcement learning to identify SMEFT operators explaining particle physics anomalies, reproducing and improving known CDF W-mass results.

The paper introduces a reinforcement learning method to search the large Standard Model Effective Field Theory (SMEFT) operator space for explanations of measurement anomalies. It was validated on the CDF W-mass anomaly, reproducing and improving known results, then applied to a harder multi-anomaly scenario. RL efficiently navigates complex loop-level operator correlations that bias human-driven phenomenological analysis.

arXiv cs.AI / cs.LG / cs.CL·Sep 9, 16:07 UTC · 7d agoAI research

22

ReactHuman: A Physics-Grounded Benchmark for Human-Like Reactive Decision-Making in Embodied Multimodal LLMs

ReactHuman benchmark tests whether multimodal LLMs react safely to sudden household hazards; seven evaluated models mishandle roughly one hazard in three.

ReactHuman is the first physics-grounded benchmark for human-like reactive decision-making, placing a multimodal LLM as the brain of a simulated humanoid facing 17 event families of sudden household hazards across over 1,000 bit-for-bit reproducible scenes with annotation-free ground truth from 240 Hz rigid-body simulation, including adversarial objects whose appearance contradicts their physics. A five-metric suite scores each reaction along reasonable, safe, and physically grounded axes, and every committed plan is physically executed. Seven representative MLLMs mishandle roughly one hazard in three, act from fixed dispositions rather than the observed scene, trust appearance over motion, and miss interception points at meter scale; none of these failures shrink with model scale.

Hugging Face daily papers·Sep 8, 20:00 UTC · 7d agoAI research

42

3D Point Splatting for mmWave Radar Novel View Synthesis

Researchers propose 3DPS, a differentiable point renderer for mmWave radar novel view synthesis that outperforms optical-NVS baselines by 1.7x-5.2x.

The paper introduces 3D Point Splatting (3DPS), the first differentiable point renderer for radar, derived from the solid-angle form of the radar equation with ITU-R P.2040 material models and complex phasor splatting. On six outdoor ColoRadar scenes it reaches 0.587 mean Pearson correlation on held-out range-azimuth images, between 1.7x and 5.2x the RadarSplat, Radar Fields, and DART baselines. The same optimized scene produces ADC, complex range profile, and RA outputs via standard FFT pipelines, and training takes about 3 minutes per scene on an RTX 4090.

arXiv cs.AI / cs.LG / cs.CL·Sep 10, 17:52 UTC · 5d agoAI research

28

Perturbation Probing: A New Diagnostic for the Fragility of LLM Safety

Unit 42 research shows LLM safety refusals concentrate in a thin neural layer, motivating external, multi-layered AI security controls.

Palo Alto Networks Unit 42 introduces Perturbation Probing, a diagnostic technique for measuring the fragility of LLM safety mechanisms. The research finds that safety refusal behavior is localized within a thin neural layer, implying small perturbations can undermine built-in refusals. The authors argue this motivates external, multi-layered security defenses on top of model-internal safety training.

Palo Alto Unit 42·Aug 28, 22:00 UTC · 18d agoAI safety & security

45

Us Thwarts Volt Typhoon Espionage

US thwarts an espionage operation by Chinese state-linked threat actor Volt Typhoon, according to the headline.

The headline indicates US authorities disrupted espionage activity attributed to Volt Typhoon, the Chinese state-sponsored group known for targeting critical infrastructure. No article text is available, so details on scope, victims, or method are unavailable.

Infosecurity Magazine·Aug 19, 00:00 UTC · 28d agoThreat actor in the wild

55

Chaotic Eclipse Releases Crowdstrike Falcon ZeroDay FalconFlank

Researcher Chaotic Eclipse released FalconFlank, a PoC zero-day privilege escalation exploit against CrowdStrike Falcon's Microsoft Office macro removal feature.

Security researcher Chaotic Eclipse (also known as Nightmare Eclipse) published FalconFlank, a proof-of-concept zero-day exploit for a privilege escalation flaw in CrowdStrike Falcon Sensor. It abuses the Microsoft Office file malicious macro removal remediation feature, which runs with high privileges, and works on fully updated Windows 11 25H2 and Windows Server 2025 with Falcon Phase 3 Optimal Protection. CrowdStrike says it is investigating and advises customers to disable the Microsoft Office File Suspicious Macro Removal Windows policy. The same researcher recently released zero-day PoCs against Kaspersky Endpoint Security (HardBreacher) and Avast Antivirus (PrettyPrague), the latter dumping the SAM database for a SYSTEM shell.

Security Affairs·Sep 9, 13:58 UTC · 7d agoExploit / PoC1· 1 read

72

FrostyGoop’s Zoom-In: A Closer Look into the Malware Artifacts, Behaviors and Network Communications

Unit 42 dissects FrostyGoop OT malware, which sent Modbus TCP commands to disrupt heating for 600+ apartment buildings in Ukraine.

Palo Alto Unit 42 analyzed new FrostyGoop/BUSTLEBERM samples, the ninth known ICS-centric malware, first used against a Ukrainian municipal energy company by sending Modbus TCP commands to ENCO control devices. The attack caused a two-day heating outage affecting over 600 apartment buildings during sub-zero temperatures. The Go-compiled malware accepts Modbus register operations and timing parameters via command-line arguments or a JSON configuration file, uses an obscure open-source Modbus library, and checks the Windows BeingDebugged flag to evade debuggers. Unit 42 telemetry recorded over 1 million Modbus TCP devices exposed to the internet between September and October 2024.

Palo Alto Unit 42·Aug 19, 12:04 UTC · 28d agoMalware in the wild

58

Repeat-After-Me: Black-Box Adaptive Visual Prompt Injection

Researchers unveil Repeat-After-Me, a black-box visual prompt injection achieving over 80% success on Qwen3.6-27B and 47% on GPT-5.5.

Researchers present Repeat-After-Me, a black-box adaptive visual prompt injection that induces frontier VLMs to reveal PII or make malicious tool calls via injected images. It exceeds 80% attack success rate on Qwen3.6-27B and 47% on GPT-5.5 even when the benign user prompt is unrelated and does not authorize the injected task. In a real-world OpenClaw Discord deployment, a minimally injected image can overwrite TOOLS.md, enabling later remote code execution and secret exfiltration.

arXiv cs.CR·Sep 3, 22:44 UTC · 12d agoAI safety & security

55

Analysis of Smoke Loader in New Tsunami Campaign

Fake Japanese Meteorological Agency tsunami warning emails delivered Smoke Loader and AzoRult malware to steal credentials from targets in Japan.

A fake tsunami warning email impersonating Japan's Meteorological Agency asked recipients to click a link on a registered fake agency domain, delivering the commodity loader Smoke Loader to targets in Japan. Smoke Loader, active since 2011, is modular, and its payloads have included banking trojans, ransomware, cryptominers, password stealers, and PoS malware; the campaign later also deployed AzoRult. New samples add junk-jump obfuscation, encrypted network traffic and payload files, a unique machine ID used for tracking and encryption, and PROPagate injection into explorer.exe, with persistence via a Startup folder shortcut and RC4-encrypted C2 communication.

Palo Alto Unit 42·Aug 17, 12:54 UTC · Aug 17, 2026Malware in the wild

45

WarmBloodAban/Minimax-h3_Singularity — new model trending #22 on Hugging Face

Community fine-tune Minimax-h3_Singularity enhances MiniMax-H3 video generation with HDR quality, distant face restoration, and improved motion, trending #22 on Hugging Face.

Minimax-h3_Singularity is a community fusion fine-tune of the MiniMax-H3 multimodal video generation model, built from multiple checkpoints and refined with pruning and weight optimization. It supports Text-to-Video, Image-to-Video, Reference-to-Video, and Video-to-Video workflows in ComfyUI, and claims improvements in HDR clarity, distant face restoration, motion fluidity, and fantasy VFX. The authors recommend pairing it with the minimax_h3_ref2v_turbo_4step_v0.1 LoRA for four-step accelerated inference, and an online demo is available via RunningHub.

Hugging Face trending models·Sep 5, 18:51 UTC · 10d agoModel release7· 1 read

28

We have a year to fix security everywhere

Blog post warns that cheap open-weight GLM 5.3-flash, once abliterated, could enable mass AI-driven vulnerability exploitation, urging industry-wide patching now.

An essay argues that Z.ai's open-weight GLM 5.3-flash—runnable locally on roughly $6k consumer hardware at 20-45 tokens/second—combined with 'abliterated' variants from groups like DeAlignAI that score 0% on HarmBench-320 puts dangerous hacking capability in nearly anyone's hands. GLM 5.3 scores 84.5% on CyberGym and 54.4% on ExploitBench, versus GPT-6 Astra's 100% and GPT-5.6 Sol's 78.5%, and the author cites evidence of frontier models exploiting real-world infrastructure. The author calls for using LLMs (Project Glasswing, Daybreak) to find and fix vulnerabilities industry-wide before adversaries weaponize cheap open models.

Hacker News · security·Sep 8, 04:48 UTC · 8d agoAI safety & securityHN 278↑ · 292 comments

55

Injected and Leaked: Actively Inducing Side-Channel Leakage Using Electromagnetic Injection and Hardware Nonlinearity

Researchers introduce InjectEave, using electromagnetic injection and hardware nonlinearity to induce side-channel leakage and eavesdrop on headphone audio from 30 meters.

An arXiv paper shows electromagnetic injection can actively amplify side-channel leakage: nonlinear hardware such as amplifiers, ADCs, and power converters modulates secret electrical signals onto an injected EM carrier, upconverting low-frequency secrets into measurable EM emissions. By tuning injection frequency and amplitude, an adversary can shape the effective spectrum and entropy of the resulting leakage. The InjectEave attack demonstrated eavesdropping on wired and wireless headphone audio from up to 30 meters and in through-wall scenarios using accessible RF equipment, plus leakage of smart home device power consumption and analog sensor inputs. Case studies show closed-loop eavesdropping and manipulation of landline phone conversations, and the paper discusses mitigations.

arXiv cs.CR·Sep 4, 06:22 UTC · 12d agoResearch

34