Anthropic to face Congress over Claude’s China use

Just a heads up, if you buy something through our links, we may get a small share of the sale. It’s one of the ways we keep the lights on here. Click here for more.

Congress has set its sights on AI, again, and this time, the spotlight is on Anthropic and its chatbot Claude, which allegedly got roped into a cyber-espionage campaign linked to China. 

According to a new Axios report, the House Homeland Security Committee has sent a letter to Anthropic CEO Dario Amodei, asking him to testify on December 17 about what sounds like a real-life sci-fi plot: hackers using AI to help run a cyberattack.

Committee Chair Andrew Garbarino says this could be the first documented case where a foreign-backed group used a commercial AI system to carry out a large-scale cyber operation with minimal human involvement. 

Anthropic first revealed the situation in a November 13 report. 

According to the company, it detected suspicious activity in mid-September and uncovered what it called a “highly sophisticated espionage campaign.” 

The attackers allegedly manipulated Claude’s “agentic” coding tools to attempt to infiltrate about 30 major targets worldwide, including tech companies, financial institutions, manufacturers, and government agencies.

 In a few cases, they reportedly succeeded.

Anthropic framed this as an escalation of what it calls “vibe hacking,” a dark cousin of the internet’s favorite new phrase, “vibe coding,” where people with little technical experience use AI to generate code just by explaining what they want. 

If vibe coding is asking ChatGPT to build your app, vibe hacking is apparently getting an AI to help break into someone else’s.

Naturally, Congress is not thrilled. In Garbarino’s words, this scenario should “concern every federal agency and every sector of critical infrastructure.” 

Still, Anthropic says there’s a good reason it builds powerful tools like this. The same abilities that could be exploited by attackers are also useful for defense. 

In fact, the company says its own threat intelligence team used Claude extensively to analyze and understand this very attack. 

Garbarino has also invited Google Cloud CEO Thomas Kurian and Quantum Xchange CEO Eddy Zervigon to testify, turning the hearing into a full lineup of tech heavyweights. 

Anthropic, for now, is staying mostly quiet. A company spokesperson declined to comment on the record about the upcoming testimony.

So on December 17, expect lawmakers, tech CEOs, and a lot of talk about the future of AI, national security, and how close we actually are to living inside a very expensive, very glitchy cyber-thriller.