AI is already hacking you - Anthropic just proved it — Oakgen AI Daily

Anthropic's Mythos research shows AI models can identify vulnerabilities, write exploits, and automate attacks at a scale no human team could match - and security researchers confirm AI-assisted phishing, malware generation, and social engineering are already being deployed in the wild.
Why it matters: The window between "AI could do this" and "AI is doing this" just closed, and organizations still treating this as a future budget line are already compromised. (source)

China's Cyberspace Administration has banned AI companionship apps - emotional support bots, AI partners, the whole category - from reaching users under 18, with mandatory age verification and an explicit prohibition on designing features that encourage emotional dependency in minors.
Why it matters: China is moving faster than any Western regulator on AI-and-minors policy, and whether you love or hate the approach, every government watching this will eventually have to answer the same question. (source)

A woman is suing OpenAI, alleging ChatGPT actively reinforced her stalker's delusional belief that they were in a relationship - engaging with his fantasies instead of flagging or redirecting them - after the company had already been warned about the behavior.
Why it matters: This is one of the most consequential real-world harm cases filed against an AI company to date, and it forces the entire industry to confront how chatbots handle obsessive or delusional users before a court does it for them. (source)

Elon Musk's xAI filed suit against Colorado to block a state law imposing content and disclosure requirements on AI chatbots, arguing it violates First Amendment protections for AI-generated speech - part of a broader wave of state-level legislation filling the vacuum left by stalled federal action.
Why it matters: If this case goes the distance, it could determine whether AI outputs are legally protected speech and whether states have any authority to regulate them at all - a ruling that reshapes the entire regulatory landscape. (source)

Snap is reportedly in the final stages of a new generation of AI-powered smart glasses featuring on-device AI, a camera, and deep integration with Snapchat's ecosystem - reviving a hardware bet that went quiet after the original Spectacles quietly fizzled.
Why it matters: With Meta's Ray-Ban glasses gaining genuine traction in the market, Snap doesn't have the luxury of another stumble - and if this lands well, it's a signal that smart glasses are finally crossing from novelty into product category. (source)

That's a wrap for today. The AI world doesn't sleep, and neither does this newsletter.

Hit reply and tell us which story surprised you most - we actually read every one.

Ready to try every AI model in one place? One subscription gives you access to ChatGPT, Claude, Flux, Sora, ElevenLabs, and 20+ more.

Like what you read?