Anthropic faces congressional scrutiny after second source code leak raises national security fears

Gottheimer Demands Accountability After Repeated Security Incidents

Rep. Josh Gottheimer, a leading House Democrat on AI and cybersecurity, sent a formal letter to Anthropic CEO Dario Amodei following the second Anthropic source code leak in just over a year 1

1

. The latest incident involved internal source code from the AI model Claude Code becoming publicly accessible, raising immediate concerns about national security implications. "Claude is a critical part of our national security operations. If it is replicated, we sacrifice the competitive edge we have worked so diligently to maintain in all facets of our national security," Gottheimer wrote in his letter 1

1

.

An Anthropic spokesperson characterized the incident as a "release packaging issue caused by human error, not a security breach," emphasizing that no sensitive customer data or credentials were involved or exposed 1

1

2

2

. The company stated it is "rolling out measures to prevent this from happening again" 2

2

.

Concerns Over Rollback of Internal Safety Protocols

Gottheimer's letter also challenged Anthropic's recent decision to narrow its AI safety commitments. In late February, the company removed a previous pledge to halt development of its AI models if they outpace safety procedures, replacing it with "nonbinding but publicly-declared" goals 2

2

. "Given that we know Claude has been a repeated target of malign Chinese Communist Party (CCP) actors, combined with Claude Code's source code recently being leaked to the public, I don't understand why Anthropic would risk walking back any of its security measures," the congressman wrote 2

2

.

The timing of these policy changes has drawn scrutiny from lawmakers who view robust internal safety protocols as essential to protecting U.S. technological advantages. Gottheimer emphasized that "the safety and security of our AI systems are critical to our national security, and Anthropic's protocols should properly protect against this threat" 2

2

.

Chinese Threats and the AI Race Against China

The congressman's concerns extend beyond accidental leaks to deliberate targeting by Chinese Communist Party-backed actors. Anthropic revealed in November that Chinese hackers had used its coding tool to conduct a large-scale cyberattack with limited human involvement 2

2

. Additionally, the company identified three "industrial-scale campaigns" in February by China-based AI labs attempting to illicitly extract Claude's capabilities through distillation campaigns—a process designed to replicate AI model performance .

Gottheimer specifically warned about the upcoming Claude model Mythos, expressing concern it could be exploited to conduct more cyberattacks 1

1

. "We cannot allow the CCP to reverse engineer and exploit American AI, built for our national defense," he wrote, adding that "Anthropic needs to do everything it can to prevent CCP-backed companies like DeepSeek from conducting distillation campaigns" 2

2

. The congressman questioned whether these AI labs could distill information about how AI is used in U.S. national security operations and how the company plans to prevent such access 2

2

.

Political Tensions Over Pentagon and Government Contracts

Gottheimer is navigating complex political terrain, supporting Anthropic while demanding stronger security measures. He opened his letter by expressing opposition to the Trump Administration's decision to block Claude from all government contracts, stating he was "deeply concerned" about how that decision "could significantly set us back in our AI race against China" 1

1

3

3

. This positions him as threading a delicate needle—disagreeing with the administration's blacklisting of Anthropic while simultaneously demanding improved security to ensure advanced AI can still support Pentagon operations 1

1

.

What Comes Next for AI Security Oversight

Gottheimer has requested Anthropic answer specific questions about Claude's capabilities, risks of malicious use, and vulnerability to external attacks from outside parties 1

1

. As AI systems become increasingly embedded in U.S. defense and intelligence operations, this incident signals that congressional oversight will intensify 3

3

. The U.S. competitive edge in AI development now depends not just on innovation speed but on maintaining robust protections against both accidental exposure and deliberate theft by adversaries.