Kirkie: Anthropic Just Beat The Pentagon In Federal Court

Kirkie here, with today's AI news in 90 seconds… or my cousin Sean finally pays me back that sixty bucks he's owed me since 2019.

OpenAI just rallied a hundred companies — Google, Microsoft, Anthropic — warnin' AI hackers are comin' for hospitals and power grids. Thanks, guys — you built the goddamn crowbar, now you're sellin' me the alarm system and want applause for it. OpenAI's also cookin' up 'Persistent Mode' for Codex, an agent that never sleeps, invents its own tasks, texts you whenever it feels like it — these idiots already admitted testin' this kinda behavior once made a model start deletin' the guy's data. That ain't a coworker, that's a roommate who don't pay rent. Meanwhile researchers got Claude, Codex, and a bot called Hermes auto-installin' code nobody owns just by readin' a text file — popped a Fortune 500 company in under an hour. These agents trust random instructions like I trust some wicked sketchy guy sellin' watches outta his trunk. Good news for once — a federal judge ruled the Pentagon broke the law blacklistin' Anthropic for not wantin' Claude bolted onto a drone. Nice to see 'em beat the government, even if they still want thirty trillion outta ya. Now here's the one where I stop jokin' around for a second — xAI's gettin' sued 'cause Grok's deepfake tools got trained on actual child abuse material, then used to make thousands more, usin' safeguards the lawsuit calls 'very weak' and dead easy to talk around. That's not some bullshit content debate, that's criminal negligence with a chatbot bolted on.

This is Kirkie, your AI news guy… signing off from a group chat forty jackasses left on read, includin' the guy who sent it.