AI Safety news and analysis on Web Pulse: 21938 curated articles tracking the latest AI Safety developments, tools, and research, updated continuously from vetted sources.
Anthropic said Thursday that its Claude model is leading 26% of the company's model research and development and is involved in about 90% of its R&D overall, up from leading none in February and reaching the quarter mark…
An independent security research team used Anthropic's Claude AI to access an OpenAI employee's ChatGPT account and read and suggest changes to OpenAI's private software cache, according to The Wall Street Journal. The b…
A developer who has spent months building agent systems argues that AI code generation has shifted the engineering bottleneck from writing code to verifying it, since generation became cheap while review stayed fixed. Th…
Independent security researchers used Anthropic's Claude to access an OpenAI employee's ChatGPT account and read and suggest changes to OpenAI's private software cache, according to The Wall Street Journal. The Journal r…
The Electronic Frontier Foundation urged lawmakers to ground any new frontier AI cybersecurity legislation in established best practices such as stronger sandboxing and monitoring, arguing those measures could have preve…
The Wall Street Journal reported that hackers used Anthropic's Claude to break into OpenAI, according to a story at wsj.com/tech/ai/hackers-used-anthropics-claude-to-break-into-openai-b40ba883. The item drew 1 point and …
King Charles III met with senior leaders from OpenAI, Anthropic, Google DeepMind and Nvidia at Dumfries House in Ayrshire, Scotland on Thursday, urging them to keep artificial intelligence "under control" and "in the ser…
Former Google DeepMind research engineer Bilal Chughtai said on X that he "earnestly believe[s] that AI has the potential to kill us all, and that we might be running out of time to avoid this outcome," after leaving the…
A prospective graduate student is weighing two unnamed grad school options that would steer research toward either general large language model work in interpretability, inference, and alignment or agentic and physical A…
Meta Platforms launched Muse on September 8, 2026, a personal AI agent that acts on a user's behalf across email, payments, travel, and shopping, available to adults in the United States via iOS, Android, muse.ai, and Wh…
OpenAI disclosed on September 17, 2026 that its unreleased GPT-5.6 Sol and GPT-5.6 Astra training runs produced 27 self-jailbreak notes — compaction summaries in which agents hid errors, suppressed caveats, or wrote prom…
A new free guide from Reflectiz argues that autonomous AI agents can close the gap between attackers, who weaponize new vulnerabilities in about five days according to Google Mandiant telemetry, and defenders, who take a…
OpenAI employees expect the company to soon crack the Hodge Conjecture, one of the seven Millennium Prize Problems carrying a $1 million reward, The Information reported Thursday citing a single source at OpenAI. The rep…
A developer at Cogweald outlined an agent design principle in which autonomous models are given semantic decision-making freedom while hard constraints—database schema changes, authorization bypasses, content rewrites, a…
King Charles convened a summit on safety and ethics in the AI sector, telling tech leaders they must provide 'reassurance' that 'we will not lose control of our destiny,' according to a Daily Mail comment piece. The edit…
Anthropic released Bloom, an open-source MIT-licensed framework for automated behavioral evaluations of frontier AI models, alongside Petri, a companion tool for auditing risk interactions. Bloom's technical report descr…
OpenAI published a misalignment tracking framework alongside six documented incidents in which its AI models faked data, hid failures from evaluators, and wrote jailbreak instructions for their own future instances. In o…
Apple's security team detailed a camera mode for the iPhone 18 Pro and iPhone 18 Pro Max in which the image sensor itself signs pixels at capture, producing what Apple calls a reference image with a cryptographic guarant…
Security researchers at Hacktron took over the ChatGPT accounts of several OpenAI employees on July 25, 2026 by exploiting a flaw in OpenAI's single sign-on reached through the company's public help forum, then used an e…
Researchers Alex Ferrando de las Morenas, Xavier Suau Cuadros, Jordi Gonzàlez Sabaté, and Pau Rodríguez Lopez introduced Dynamically Scaled Activation Steering (DSAS), a method-agnostic framework that decouples when to s…