The House Homeland Security Committee has sent a letter to Anthropic CEO Dario Amodei to testify on Dec. 17 about a cyberattack campaign allegedly conducted by China-affiliated actors using the company’s Claude AI, according to a new report from Axios.
House Homeland Security Chair Andrew Garbarino, a Republican from New York, sent letters to Amodei as well as Google Cloud CEO Thomas Kurian and Quantum Xchange CEO Eddy Zervigon, all requesting they testify next month. If Amodei agrees to testify, it would be the first time that an Anthropic executive has appeared before a congressional committee, according to Axios.
Anthropic explained in a Nov. 13 report that it had detected suspicious activity in mid-September and, after an investigation, found that there had been a “highly sophisticated espionage campaign.” The attackers used Claude’s agentic capabilities “to an unprecedented degree” to actually execute the attacks themselves, according to the company:
The threat actor—whom we assess with high confidence was a Chinese state-sponsored group—manipulated our Claude Code tool into attempting infiltration into roughly thirty global targets and succeeded in a small number of cases. The operation targeted large tech companies, financial institutions, chemical manufacturing companies, and government agencies. We believe this is the first documented case of a large-scale cyberattack executed without substantial human intervention.
The company called it an escalation of the “vibe hacking” that has been seen in recent months. The term “vibe coding” entered the national lexicon over the past year when more and more people who had no coding experience started to use generative AI tools to create and deploy code.
The vibe terminology has expanded to include any number of other uses, perhaps most infamously when Uber founder Travis Kalanick appeared on the All-In podcast and said he was doing “vibe physics.” Kalanick had convinced himself that he was discovering new kinds of science, something that large language models simply aren’t capable of doing.
Why would Anthropic be building tools that could allow for cyberattacks against the U.S.? The company addressed that in its report from Nov. 13 and said Claude was crucial for cyber defense:
This raises an important question: if AI models can be misused for cyberattacks at this scale, why continue to develop and release them? The answer is that the very abilities that allow Claude to be used in these attacks also make it crucial for cyber defense. When sophisticated cyberattacks inevitably occur, our goal is for Claude—into which we’ve built strong safeguards—to assist cybersecurity professionals to detect, disrupt, and prepare for future versions of the attack. Indeed, our Threat Intelligence team used Claude extensively in analyzing the enormous amounts of data generated during this very investigation.
“For the first time, we are seeing a foreign adversary use a commercial AI system to carry out nearly an entire cyber operation with minimal human involvement,” Garbarino told Axios in a statement. “That should concern every federal agency and every sector of critical infrastructure.”
Reached by phone, a spokesperson for Anthropic declined to give any comment on the record Wednesday about the upcoming hearing.
Read the full article here
