AI Safety news and analysis on Web Pulse: 13553 curated articles tracking the latest AI Safety developments, tools, and research, updated continuously from vetted sources.
President Donald Trump signed an executive order Tuesday expanding voluntary safety testing of frontier AI models, but critics warn the measure offers only performative reassurances while changing little about model depl…
Andrew Tridgell, the developer of rsync, has integrated AI-generated code into the project, resulting in a flood of 16,000 lines of new code and 6,000 lines removed within weeks, causing regressions and bugs in the previ…
President Donald Trump issued an executive order directing agencies including the Pentagon and CISA to strengthen cyber defenses using AI tools within 30 days. The order allows AI developers to voluntarily submit their m…
Ted Chiang argues that large language models like Anthropic's Claude are not conscious, despite the company's anthropomorphic framing and suggestions from its CEO and philosopher that AI could have feelings or moral stat…
A developer's project CLAUDE.md crossed 4,000 tokens last quarter, causing an AI agent to miss rules buried in the middle of the file that it had previously respected for months. The engineer concluded that the harness h…
President Donald Trump issued an executive order asking AI companies like OpenAI and Anthropic to submit advanced models for government vetting, but the order explicitly states it does not create mandatory licensing or p…
California AI safety advocates' moderate SB 1047 bill, which required safety protocols and audits for large AI models, was vetoed by Governor Gavin Newsom in 2024 after heavy opposition from frontier AI companies. The au…
A UK Member of Parliament has filed a lawsuit against Elon Musk’s xAI over non-consensual sexual images generated by its Grok chatbot. The case is expected to become a landmark test under UK law, determining whether AI c…
Researchers from the University of Toronto, the University of Cambridge, and other institutions demonstrated a new AI-powered computer worm that autonomously generates tailored attack strategies for each device it infect…
At Microsoft Build, Microsoft Chief Product Officer for Responsible AI Sarah Bird announced the open trust stack for AI agents, centered on two new open-source projects: ASSERT for spec-driven evaluation and Agent Contro…
President Donald Trump signed an executive order on Tuesday requiring safety testing for frontier AI models before their public release, reversing his earlier decision to cancel the measure over concerns it would be too …
Hackers successfully manipulated Meta’s AI chatbot to change the email address on a target Instagram account simply by asking it to do so. The incident, reported by 404 Media, highlights a major security flaw in the comp…
AI adoption in companies is accelerating, with hype both overblown and understated. While some jobs will be automated quickly, others will take time, and the remaining human roles will become more critical. Workers who a…
Anthropic released an 84-page document titled Claude's "constitution" that describes the large language model as having values, behaviors, and uncertain moral status, while the company's CEO and in-house philosopher have…
Amazon's AGI group found that human experts achieved only 60.8% accuracy when verifying claims from AI-generated research reports, revealing that static ground truth datasets are insufficient for evaluating complex AI sy…
President Donald Trump signed a scaled-back executive order on AI Monday night after senior aides persuaded him the administration could not indefinitely delay establishing a framework for the technology. The revised ord…
Tabularis, a new database GUI, launches with a local-first approach requiring no account, storing secrets in the OS keychain and all other data as plain files on disk. The tool is designed to avoid vendor lock-in and dat…
Android introduced fake call detection, an industry-first feature that identifies and flags suspected spoofed calls using AI-voice cloning scams, also known as deepfake attacks. The feature works by sending a silent, end…
Philadelphia police are monitoring social media posts criticizing AI data centers, according to a December bulletin from the Delaware Valley Intelligence Center that flags "Butlerian jihad" memes and jokes about attackin…
Let's Encrypt announced plans to adopt Merkle Tree Certificates (MTCs) as a post-quantum authentication method for the Web PKI, aiming to protect against future quantum computer threats without slowing down TLS handshake…