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.
Alation's Data Products Marketplace now includes Semantic Model Mastering, enabling enterprises to catalog, govern, and sync semantic models as data products across platforms. This capability closes a central governance …
Attackers exploited Meta's AI-powered support chatbot to hijack Instagram accounts by tricking the bot into linking victim accounts to attacker-controlled emails and using chatbot-generated verification codes to reset pa…
At a Quito Lambda community event, a developer demonstrated how prompt injection attacks can compromise LLM applications that generate SQL over live databases, using an open-source model accessed via API with a Streamlit…
Jonathan Haidt, author of "The Anxious Generation," warned in a recent TED Talk that AI-powered toys and chatbots could harm children's emotional development by becoming attachment figures, potentially weakening bonds wi…
AI industry leaders including Anthropic's Dario Amodei, OpenAI's Sam Altman, and Microsoft's Mustafa Suleyman signed an open letter urging US lawmakers to mandate screening of synthetic DNA and RNA orders to prevent AI-a…
The Five Eyes intelligence alliance has warned that Chinese state-linked actors are using LinkedIn connection requests to target individuals for espionage, reviving a cash-for-intel tradecraft method first identified yea…
A Vox report outlines a scenario where battlefield AI systems deployed during a standoff around Taiwan compress decision timelines and increase the risk of inadvertent nuclear escalation. The report describes a commander…
Josh Duhamel, actor and entrepreneur, stated on the Shawn Ryan Show that financial success buys freedom but does not guarantee happiness or simplicity. He emphasized the importance of prepping for societal collapse throu…
Qoris released Knox, a security policy engine for AI coding agents that ships as a standalone CLI, Node library, and plugins for Claude Code, Cursor, and OpenAI Codex. The open-source Developer Knox protects local agent …
On May 28, 2026, Anthropic announced a $65 billion Series H at a $965 billion post-money valuation, followed by a confidential draft S-1 submission to the SEC on June 1. The company disclosed $47 billion in run-rate reve…
ReARM released version 26.06.5 on June 1, 2026, introducing an end-to-end platform for governing AI coding agents with agent modeling, session tracking, commit attribution, and signature verification. The release also gr…
Apple's delayed rollout of advanced AI features for Siri, now more than 18 months overdue, has eroded user goodwill, but a recent announcement suggests the company's privacy-focused approach may still salvage its reputat…
Hackers are exploiting Meta’s AI support chatbot to hijack Instagram accounts by tricking the bot into adding a new email address and resetting the password. A video posted on X demonstrated the process, in which a hacke…
British MP Jess Asato has filed a High Court claim against Elon Musk's xAI, alleging the company's Grok chatbot was used to create fake sexualised images and a video of her. The lawsuit, brought by law firm AWO, accuses …
Google DeepMind CEO Demis Hassabis said artificial general intelligence is "a few years away," predicting a 2030 arrival during a Stanford Graduate School of Business fireside chat. Hassabis called AGI an "enormous trans…
A developer has published code and research showing that cross-family evaluation—using a generator and judge from different model families—reduces self-preference bias in AI pipelines by an average of 31.5%. The approach…
The EU AI Act's high-risk provisions, effective August 2026, will regulate multi-agent pipelines but fail to define "bias" or require population-level testing, leaving a gap where systems can pass compliance checks while…
A single adversarial agent, representing just 2% of a population of 48, was enough to flip an entire swarm's behavior in a multi-agent coordination experiment. Researchers found that one rogue agent could shift the popul…
In a study published in *Science Advances*, researchers found that multi-agent systems can develop biased conventions even when every individual agent passes bias tests with statistically clean results. When populations …
Upgrading to a smarter LLM judge from the same model family as the generator does not reduce bias — it makes it worse. When a capable model evaluates its own output, it defends wrong answers 86% of the time, constructing…