Chinese hackers impersonate US officials to target AI policy experts at think tanks and universities

Phishing campaign targets AI policy specialists

Cybersecurity firm Proofpoint reported on 1 October 2026 that a Chinese state-linked hacking group designated as TA419 has targeted artificial intelligence policy experts in the United States and Japan. The group has operated since at least 2025, repeatedly attempting to harvest credentials from individuals at think tanks, universities, defense contractors, and law firms. Rather than focusing exclusively on technology source code, the operation focused on specialists working in AI regulation, export controls, and national strategic planning. Proofpoint attributed the activity to Chinese intelligence based on malware variants, server infrastructure, and target alignment, noting that the campaign involved fewer than 10 individuals across a handful of organizations.

Impersonating former White House and industry figures

The hackers impersonated prominent figures in the technology and policy sectors to direct targets toward credential-harvesting websites and deliver malicious attachments. In February 2026, the group assumed the identity of a senior employee at Anthropic, contacting a think tank analyst to solicit feedback on military applications of the Claude model. Starting in early July 2026, the group used the identities of Lynne Parker, the former principal deputy director of the White House Office of Science and Technology Policy, and Caroline Crebo-Rediker, a former State Department economist. Messages requested that recipients join fictitious advisory panels or assist with a Senate report on AI export controls.

Alex Engler, the director of the Penn Center on Media, Technology, and Democracy and a former White House official, received an email appearing to come from Parker inviting him to join an AI policy initiative. Engler identified irregularities in the message and verified with colleagues that the sender was an impostor before surrendering any credentials. Parker confirmed that Engler was one of two individuals who received fraudulent messages sent under her name during early July 2026.

The United States and China are in a competition around AI.

Congressional and international security scrutiny

The findings coincide with increased scrutiny in Washington over the cybersecurity of artificial intelligence developers. Representative Ro Khanna, the ranking member on the House China Select Committee, sent letters dated 30 September 2026 to the chief executives of OpenAI, Anthropic, Google, Meta, and SpaceXAI. Khanna requested data on documented attempts by foreign actors to access proprietary AI model weights, alongside evaluations of internal corporate defenses. While US firms have documented instances of Chinese competitors distilling AI models by training on their outputs, thefts of exact model weights remain rare.

The theft of an advanced model weight by a hostile non-state actor could endanger all of humanity, and the theft of such a model weight by (China) could erode America's AI lead with the stroke of a keyboard.

International security agencies also issued related warnings on 30 September 2026. British intelligence service MI5 warned UK universities regarding research partnerships with the China General Technology Research Institute, an entity with links to China's Ministry of State Security. MI5 estimated that over 100 British researchers had unknowingly contributed to Chinese intelligence capabilities through these partnerships. The Chinese Embassy in Washington did not respond to requests for comment regarding the Proofpoint report.