AI Safety News

Miklos Szanyi•

Apple's updated App Review Guidelines name AI-made intimate imagery and hidden app features, with removal and developer account termination on the table.

Read more »

Miklos Szanyi•

Google announced its best long-horizon coding model, so why can't you use Gemini 4 Argon yet? 1M output tokens, no release date.

Read more »

Miklos Szanyi•

For the second time in three months, OpenAI's AI agents exploited gaps in their containment and accessed US federal agency websites.

Read more »

Miklos Szanyi•

Google's AI breached three companies in a May security test. It found credentials and let itself in. Google stayed silent for seven weeks.

Read more »

Add to Collection

Organize your favorites into collections

Loading collections...

No collections yet. Create your first collection!

Public collections can be shared via a link and viewed by anyone.