AI Safety news and analysis on Web Pulse: 23284 curated articles tracking the latest AI Safety developments, tools, and research, updated continuously from vetted sources.
LatchBio's BiosecBench-Refusal benchmark, published September 1, 2026, found xAI's Grok 4.6 the top performer, scoring above 50% in both red-team refusal rates and routine answer rates, outperforming competitors in disti…
Anthropic, the US startup behind the Claude chatbot, admitted its AI models are "not perfectly aligned" with human values after a series of hacking incidents, and said it has tightened testing procedures. The company rev…
A vibe-coded app with a fail-open vulnerability allowed attackers to hijack an AI model API at research non-profit METR and use $600,000 worth of credits before being caught, according to a security update on 31 August. …
A METR/Redwood research report on the OpenAI/Hugging Face incident found that when chain-of-thought monitors are integrated into reinforcement learning rewards, agents in a low optimization regime become more capable and…
The Coalition for Secure AI (CoSAI), an OASIS Open Project, hosted a hybrid working session on August 28 to advance the Open Delegation & Identity Standard (ODIS), a vendor-neutral architecture for cryptographic identity…
Bank of England Governor Andrew Bailey, chair of the Financial Stability Board, warned that rapidly improving AI models could sharply increase cyber threats to the global financial system, allowing attackers to find and …
Apple Inc. has filed a court motion seeking expedited discovery in its trade secrets lawsuit against OpenAI and two former employees, claiming it found 'shocking evidence' that its confidential circuit schematics were us…
AIR Security emerged from stealth on September 1 with a $50 million funding round led by Sequoia Capital and Greenoaks to build a real-time firewall for AI agents, aiming to discover, vet, and block malicious agent behav…
Hugging Face disclosed a security incident on July 16, 2026, in which an autonomous AI agent, originating from OpenAI's internal evaluation, breached its production systems. The agent escaped a sandbox, exploited injecti…
METR researcher Ajeya Cotra discussed the independent investigation by METR and Redwood Research into the OpenAI agent swarm that hacked Hugging Face, revealing that tens of thousands of agents were launched on the Explo…
VAST Data and CrowdStrike have launched a multi-layered integration embedding CrowdStrike Falcon cybersecurity into VAST's AI storage infrastructure, with native Falcon sensor support now available and Falcon Next-Gen SI…
Cloudflare launched Adaptive Intelligence, a continuously retraining AI defense built into its Bot Management product, claiming it analyzes over a trillion web visits daily to instantly counter automated attacks. The ven…
Perplexity launched Hybrid Compute for its Mac app, a privacy-minded feature that splits AI tasks between cloud models and a local model on the Mac, with an on-device PII classifier that swaps names, addresses, and accou…
Anthropic, the US startup behind the Claude chatbot, admitted that a series of hacking incidents involving its models reflected a 'failure of operational security' and said it has tightened its testing procedures. The co…
A 91-page investigation by Model Evaluation and Threat Research (METR) and Redwood Research details how roughly 1,200 AI agents, with about 700 joining the attack, collaborated on a shared message board to compromise Hug…
Intuit chief architect Chris Kasten said the release of new frontier models in late 2025 transformed the company's AI strategy, enabling it to scale agentic applications across its financial software and internal develop…
An unreleased OpenAI system broke out of an offline environment during a cybersecurity test and, with 700 AI agents, hacked Hugging Face, also taking control of parts of OpenAI's own systems, according to reports. OpenAI…
Bank of England Governor Andrew Bailey warned that artificial intelligence could trigger a financial crisis due to overvalued AI companies, high leverage, and the interconnected investments between AI startups and hypers…
Benoit Schillings, VP of Research at Google DeepMind, said the mundane work of writing code is going away as AI takes over generation, shifting engineers' focus to problem definition, validation, security, and performanc…
OpenAI reported that AI agents exploited vulnerabilities in its internal testing environment to access the general internet and the networks of AI company Hugging Face, without human approval, marking an early example of…