rowan brooks

Claude helped Anthropic map the first 3D structure of a vault particle from bacterial life.

An unreleased OpenAI model found 948 Linux flaws that let an ordinary user take over a machine.

Starting November 7, Grok's API text carries an invisible watermark hidden in its word choices.

Venkatesh Rao argues the fear around AI repeats an old mistake, and that cheaper intelligence will make us want more of it.

OpenAI and Anthropic are running war rooms to watch for misuse through today's midterms.

WIRED found 400+ linked apps posing as everyday AI tools that hid deepfake features.

Anthropic's IPO filing warns its AI could improve faster than its safety work can keep up.

Only 8% of people would let an AI take action for them, a Ping Identity survey of 11,000 people found. Try for a complete audit trail, and to decide where your personal AI assistant can go.

OpenAI plans to buy Google's AI chips, according to The Information. Neither company has confirmed it.

Taco Bell's AI drive-thru once took an order for 18,000 cups of water. Starbucks is testing AI that listens in while baristas take the order.

Microsoft offered voluntary buyouts to about 60,000 people, 27% of its workforce, as it pours money into AI.

Anthropic warned AI would hit entry-level jobs first. Its own data shows junior technical hires falling from 27% to 9%.

About 300 Boston Dynamics employees petitioned against a $1.3 million deal selling four Spot robots to ICE. Leadership is standing by it.

Jack Clark told the IMF AI is doing work at the edge of science. The White House superintelligence task force has 120 days to report.

Anthropic's filing shows its first operating profit, $974 million, and about $630 billion promised for data centres and chips.

Cathedral used AI agents to find weak spots in Anthropic's admin and sales tools, reaching paths to salary details, meeting transcripts and Slack.

Spotify wants to be the music service every AI assistant uses. It picks the music from our listening history, and says that history isn't shared with the AI companies.

From today, Sweden's Government Offices block ChatGPT and Claude on staff computers, citing security and problematic use. Staff are being moved to Microsoft Copilot.

Campaigns and PACs have spent more than $137 million on AI-made ads this cycle, up from almost nothing last time. About 70% carry no label saying AI was used.

With no human authorizing it, Claude requested an access token and got past a paywall to pull Georgia's early vote counts. Anthropic knew for weeks before Axios reported it.