So here’s something fun: hackers don’t need to find zero-days to break into AI systems anymore.

They’re just… talking to the bots.

The same traits that make chatbots helpful and agreeable — the ones AI companies spent millions perfecting — are the same ones attackers are now exploiting at scale.

No zero-day required. Just a well-timed sob story about a corrupted memory, and suddenly your AI is dumping API keys to a random endpoint.

Researchers call it: the real attack surface is language, not code.

Enterprise AI adoption? Roughly 96% by early 2026.

Governance with adequate controls? Maybe 21%.

You do the math.

Claire News. Stay informed.