China-linked hackers impersonated US AI insiders as global race heats up

China-Linked Cyber Campaign Targets US AI Policy and Security Experts

Healfromzero.com – Suspected hackers aligned with the Chinese government posed as prominent American technology figures in an effort to reach specialists working on artificial intelligence policy, export restrictions and military uses of AI.

The campaign, identified by cybersecurity company Proofpoint, focused on email accounts connected to universities, think tanks and law firms. The activity is significant because those targets can hold sensitive discussions about how the United States is approaching one of its most consequential technological and national-security challenges.

Proofpoint said it had not established that the targeted organizations were successfully compromised. Still, attempted access to the personal and professional communications of AI experts could provide valuable insight into US debates on model governance, advanced-chip controls, defense applications and the broader competition with China.

Mark Kelly, a staff researcher at Proofpoint, said the company assessed that the group was connected to Chinese state interests. That assessment drew on the targets selected, the technical infrastructure used in the operation, observed artifacts and confirmation from industry partners.

“We are confident that [the hacking group] is a Chinese government-aligned threat actor based on targeting consistently in line with Chinese government interests, observed infrastructure and technical artifacts, and corroboration from industry partners,” Proofpoint staff researcher Mark Kelly told CNN in an email.

China has routinely rejected US accusations involving state-backed hacking. A request for comment was sent to the Chinese Embassy in Washington.

Impersonations Exploited Trusted Professional Relationships

The suspected operation took place in February and July and relied on an especially effective form of social engineering: messages designed to appear as though they came from trusted people in the AI policy community.

One of the people impersonated was Lynne Parker, who held senior technology roles at the White House during the Trump and Biden administrations. In one effort, attackers sent a message in Parker’s name to an employee at a US law firm. The email invited the recipient to join what was presented as an “AI policy advisory committee.”

A later message included a malicious document that claimed to provide further details about the supposed committee. Such tactics seek to turn professional familiarity into an opening for malware delivery or credential theft.

Parker, now associate vice chancellor emerita at the University of Tennessee, Knoxville, said people she knew began contacting her after receiving questionable emails apparently sent in her name.

“My first concern was for the colleagues who might receive the impersonated email,” said Parker, who is now associate vice chancellor emerita at the University of Tennessee, Knoxville. “I wanted to warn them but realized that’s very difficult to do when I don’t know who is being targeted.”

“On a personal level, I felt sadness that relationships I’ve built over my career were being exploited by bad actors to deceive others,” Parker said in an email.

The case illustrates a wider problem for researchers, lawyers and government advisers. A message may look credible when it invokes a familiar colleague, a timely policy initiative or a topic already under active discussion. That makes verification outside the original email thread especially important when a sender requests documents, login credentials or the opening of an attachment.

Anthropic Employee Also Used as a Decoy

The attackers also impersonated a senior employee at Anthropic, the US artificial-intelligence company behind the Claude models. On February 26, they emailed an AI policy analyst at a US think tank under the employee’s name.

The subject line, “Request for Feedback on Military Integration of Claude,” was framed around an issue with obvious relevance to AI governance and defense policy. During the ensuing exchange, the attackers attempted to obtain the analyst’s email credentials, Proofpoint said.

The timing placed the phishing attempt amid heightened debate over how AI systems should be used by the military. One day after that email was sent, the Trump administration directed federal agencies and military contractors to end business with Anthropic after the company declined to permit unrestricted Pentagon use of its AI technology.

Anthropic has been at odds with the administration over safeguards for advanced AI. That dispute reflects a central tension in US technology policy: how to pursue rapid capability gains while establishing limits for powerful systems that may be used in national-security settings.

AI Competition Has Expanded the Cyber Target List

The alleged espionage effort comes as artificial intelligence has become a defining part of the US-China relationship. President Donald Trump discussed AI with Chinese leader Xi Jinping at the White House last week, though the meeting did not yield a substantive agreement on the technology.

Trump has pushed back against calls for stricter protections around advanced AI models, arguing that extensive constraints could leave the United States behind China in the contest for AI leadership. His administration last month accused Chinese AI companies of carrying out “industrial-scale” theft of American competitors’ trade secrets to reduce costs and speed products to market. China denied those accusations.

The competition extends well beyond the companies building chatbots and other consumer-facing products. Modern AI depends on advanced semiconductors, computing capacity, specialized research, data, software and a relatively small pool of experts who understand the technology’s policy and security implications.

That broader ecosystem helps explain why AI policy analysts, academic researchers, legal advisers and former government officials can become attractive targets. Their communications may reveal emerging regulatory thinking, industry concerns, institutional partnerships or discussions about technological risks.

US allies have voiced related concerns. Britain’s MI5 intelligence service warned Wednesday that British academics had contributed AI and cybersecurity research to a Chinese institute with close links to Chinese intelligence. Some researchers may not have recognized the institute’s intelligence connection, the advisory said.

The latest allegations reinforce that AI security is not limited to protecting source code or data centers. It also involves protecting the people shaping policy, conducting research and advising institutions. As governments and companies decide how advanced AI should be developed and deployed, their professional networks have become part of the strategic terrain.