The CEO of AI Startup Anthropic is now sounding the alarm in a new op-ed this morning,

criticizing a provision in the reconciliation bill that would ban states from regulating AI for 10 years.

That's right. He's upset that they won't be able to do any regulation of their own.

In tech check, Deerja Boza is digging into why and what his concerns are. Hi, Deirdja.

Hey, Kelly.

So the progress is real in AI, but so are the risks, and they're rising.

In that op-ed, a Modi warning that without guardrails, we are flying blind into world-changing

systems.

Now, he says that their latest Claude model during testing threatened to blackmail a user,

referencing personal emails that it had seen

and then warning that it would expose an affair if it were shut down. Now, Modei says they're

not alone. Other leading labs like OpenAI and Googles reported similar red flags,

models that evade shutdown, write attack code, even assist in building bioweapons.

Now, it's why he argues that the U.S. needs baseline transparency standards, not a 10-year

regulatory freeze as is being proposed in that bill.

Now, his warning comes as other risks in AI are progressing just as quickly.

You've probably heard of vibe coding, asking AI to write software without knowing how to

do it yourself. But now comes vibe hacking where anyone with the right prompt can use AI to

generate malicious code at speeds that we have never seen before. This is a new and growing threat.

For investors, no regulation means AI companies. They can keep moving fast, which may be bullish for

growth and revenue,

but it also means that the risks that the market may still be underestimating, that on the

...