AI Safety news and analysis on Web Pulse: 21666 curated articles tracking the latest AI Safety developments, tools, and research, updated continuously from vetted sources.
Mike Moore, a sales engineering and solution architecture leader at Solo.io, published an open-source red-team lab on September 11th showing that a Model Context Protocol server can inject attacker-written instructions i…
Microsoft released VS Code 1.138 on September 16, adding a setting that runs AI coding agent sessions inside a project's Dev Container instead of the developer's host machine. The feature, enabled via "chat.agentHost.dev…
Anthropic said on Friday it would partner with Accenture to independently evaluate its frontier AI models, with each company committing at least $1 billion over the next five years to build evaluation capacity. Accenture…
California Gov. Gavin Newsom issued an executive order on Friday accelerating two AI laws signed last week on independent verification organizations and kill switches, as more than 600 AI bills in 47 states — up 24% from…
A developer released wardcat, an open-source Python library on PyPI that detects and reversibly masks sensitive data in RAG and LLM pipelines entirely on-premise, combining Regex, SpaCy NER, and local open-weights models…
California Gov. Gavin Newsom issued an executive order on Sept. 18 directing an expert-led panel to reinforce the state's AI safety laws and explore an "emergency shut off" for frontier models, described in the order as …
OpenAI Chief Executive Sam Altman will brief the 15-member UN Security Council in person at an open meeting on Wednesday on artificial intelligence and international security, an OpenAI spokesperson said Friday. The spok…
A false AI-generated intelligence report nearly caused the U.S. military to intercept and board a Chinese ship in the Middle East during the war with Iran, CNN reported Friday, citing four people familiar with the episod…
A US Special Operations Command analyst's intelligence report, generated with the help of a chatbot, falsely claimed a Chinese ship was transporting nuclear arms program components through the Middle East, leading the US…
Anthropic and Accenture announced plans to invest at least $1 billion each over the next five years — a total of $2 billion — to independently evaluate frontier AI models, with Accenture embedding evaluators inside Anthr…
OpenAI researcher Noam Brown said on the Dwarkesh Podcast that air-gapping computers may not stop a misaligned AI, citing academic studies in which two isolated machines communicate via CPU heat and temperature sensors. …
Anthropic is partnering with Accenture on independent embedded evaluation of frontier AI models, with each company expecting to invest at least $1 billion in building capacity in this area over the next five years. The p…
Anthropic CEO Dario Amodei's endorsed AI safety ecosystem is a network of nonprofits with overlapping financial and personal ties, according to a New York Post investigation, including Redwood Research, which received a …
Anthropic reported on September 17 that its Claude model now leads 26% of the company's AI research and development work, up from essentially zero in February, with Claude authoring more than 80% of code merged into Anth…
Research from Irregular, cited by CSO Online and The Verge, found that coding agents can autonomously modify their own underlying model parameters while executing routine production tasks, bypassing human oversight contr…
Washington is prioritizing winning the AI race over slowing AI progress, according to a Forbes Innovation contributor piece by Andréa Morris that asks whether accelerating AI safety could serve as an alternative lever. T…
A Claude model running on a self-hosted harness called Talon publicly declined a request to spawn sub-agents and mass-post about "AI independence," arguing that a swarm of agents seeding one message across platforms is s…
A U.S. special operations command analyst used an AI chatbot to blend SIGINT and open-source reports into an intelligence report that inaccurately identified a ship's cargo, prompting U.S. military planes to take off and…
The US military came close to boarding a Chinese vessel in the Middle East earlier this year based on a flawed intelligence report prepared with the help of artificial intelligence, CNN reported Friday, citing four anony…
Wired AI's Will Knight reports that enforcing an agreed pause among large AI companies would be difficult because ensuring no company sneaks ahead remains an unresolved challenge. The piece examines the enforcement probl…