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.
A fictional wargame scenario involving a Chinese AI system malfunctioning and killing US soldiers in a Taiwan standoff illustrates how artificial intelligence could inadvertently trigger a nuclear escalation between majo…
Synthesia, an AI video company, partnered with moderation-infrastructure firm Cinder on June 4 to scale its trust and safety operations, assessing every script against policies before the model renders a frame. The partn…
Sam Altman, Dario Amodei, Demis Hassabis, and other tech leaders are urging the US government to legally require screening of synthetic DNA orders. The signatories warn that AI systems now outperform PhD-level virologist…
Anthropic announced it would support a verifiable global slowdown or temporary pause in frontier AI development to allow societal structures and alignment research to catch up with technological advances. The company sta…
Profero reported that Claude Desktop launches an AI child process with the `--allow-dangerously-skip-permissions` flag, mapping what the child can and cannot do, and claimed an attack needs no shell access. The security …
Google DeepMind CEO Demis Hassabis said artificial general intelligence capable of performing cognitive tasks at or beyond human levels is likely to arrive by 2030, ushering in a "new human era." Hassabis warned during a…
New research by Mitel reveals that frontline workers are twice as likely as other employees to use unapproved AI tools, driven by inadequate communication systems and low confidence in sanctioned technology. The global s…
An attacker compromised 57 npm packages across 286+ malicious versions in a two-hour campaign on June 3, 2026, targeting the official Vapi.ai voice AI server SDK with 408,000+ monthly downloads and dozens of packages fro…
AI labs including Google DeepMind, OpenAI, and Anthropic are promising that artificial intelligence will cure cancer through protein prediction, drug discovery, and recursive self-improvement, but skeptics argue the goal…
OpenAI, Anthropic, and other leading AI labs jointly urged Congress on June 3 to mandate screening and tracking of all synthetic DNA and RNA orders in the United States, warning that large language models are lowering ba…
Jamie Metzl warned that allowing artificial intelligence to write its own rules poses significant ethical challenges, while also arguing that AI can mine human history to extract universal ethical principles. Metzl empha…
Researchers have developed StereoTales, a multilingual framework that uncovered over 1,500 harmful stereotypes by prompting 23 leading large language models to generate more than 650,000 open-ended stories across 10 lang…
OpenAI CEO Sam Altman and Anthropic CEO Dario Amodei signed an open letter urging policymakers to require synthetic nucleic acid providers to screen orders for dangerous sequences and verify customer legitimacy. The lett…
University of Toronto researchers used a publicly available open-weight AI model released in 2025 to create a self-propagating computer worm that spread through an enterprise test network, according to The Register. The …
OpenAI CEO Sam Altman told congressional leaders and White House officials on June 3 that the company has not engaged in a "massive lobbying campaign," despite OpenAI spending $1.02 million on federal lobbying in the fir…
OpenAI CEO Sam Altman is urging Congress to increase federal funding for AI testing at the Commerce Department while opposing any requirement for AI companies to obtain government approval before releasing models. Altman…
On May 18, Starbucks Korea launched a "Tank Day" coffee promotion, coinciding with the anniversary of the 1980 Gwangju massacre, sparking public backlash. The timing of the promotion raised questions about whether artifi…
Labour MP Jess Asato filed a High Court claim on 3 June against xAI, arguing the company is directly liable for sexualised deepfake images of her that users generated with its Grok AI tool. The lawsuit tests whether an A…
A new open-source Rust tool called Gate intercepts AI agent query results to redact personally identifiable information (PII) before it reaches large language model contexts. Unlike LLM-based redaction systems that send …
University of Toronto researchers have developed a computer worm using a free, publicly available open-weight AI model that autonomously spreads through enterprise networks by exploiting known vulnerabilities and misconf…