Chinese Cyber Group Posed as AI Experts to Breach US Policy Figures, Report Finds

A China-aligned hacking operation spent months impersonating well-known artificial intelligence specialists in an effort to break into the email accounts of American policy thinkers, according to a newly published report from the cybersecurity company Proofpoint.

The findings, made public on Thursday, describe a sustained campaign in which operatives masqueraded as respected figures in the AI field in order to win the trust of their targets. Among the identities the hackers adopted was that of Lynne Edwards Parker, who previously served as principal deputy director of the White House Office of Science and Technology Policy.

The group behind the activity is tracked by security researchers as TA419. Proofpoint's analysts believe the operation began in April 2025 and remained active through the middle of 2026, with a particularly notable wave of activity recorded in July.

Rather than relying on technical flaws in software, the attackers leaned heavily on social engineering — the practice of manipulating people into voluntarily handing over access. Their approach began with emails crafted to look innocuous, often framed as invitations to join an "AI Policy Advisory Committee" or similar professional initiative.

Those initial messages were designed less to steal information immediately than to start a conversation. Once a recipient replied and showed interest, the hackers followed up by directing the person to a counterfeit sign-in page built to capture usernames and passwords.

Proofpoint said the technique involved generating a fake browser pop-up window that appeared on top of a legitimate website. Because the prompt looked like a routine authentication request, victims were far less likely to suspect that they were surrendering their credentials to an intruder rather than to a trusted service.

The list of intended victims was broad and strategically chosen. According to the report, the campaign reached policy analysts at think tanks, employees of defence contractors, academics at universities, and lawyers at private firms. Targets were located not only in the United States but also in Japan, suggesting the operation cast a wide net across allied capitals and research hubs.

Proofpoint declined to publicly identify the individuals whose accounts were attacked. However, the news agency Reuters was able to confirm that at least one of them was Alex Engler, a former White House official who now leads the Penn Center on Media, Technology, and Democracy at the University of Pennsylvania.

Engler told Reuters that he did receive one of the suspicious emails. Suspicious of its origins, he reached out to colleagues in the field to verify whether the message was genuine. Through those conversations, he concluded that the sender was not the person they claimed to be but an impersonator using a real expert's name and reputation as cover.

The tactic of borrowing a credible identity is a hallmark of this particular group. In February, according to the report, the same actors posed as a senior employee of Anthropic, the AI safety company, in an attempt to reach specialists involved in technology policy.

That earlier episode illustrates how the group adapts its personas to its audience. By adopting the name of someone who would plausibly be in touch with Washington's technology policy community, the hackers hoped to lower the defences of people accustomed to receiving messages from peers and former government colleagues.

The timing of the campaign is significant. AI policy has become one of the most closely watched areas of the US-China relationship, with both governments racing to shape rules around advanced models, compute, and national security. People who work on those questions often hold sensitive information about regulatory plans and internal deliberations.

Researchers who study state-linked cyber operations say intelligence services frequently pursue this kind of targeting because it offers a low-cost way to gather insight into how rival governments think. A compromised inbox at a think tank or university can yield drafts of policy papers, private correspondence with officials, and early warning of forthcoming decisions.

Proofpoint's assessment is that TA419 is unlikely to stop. The firm expects the group to keep aiming at think tanks and other organisations that shape policy, and to continue hiding behind the identities of real experts to do so.

Security professionals generally advise potential targets to verify unexpected messages through a separate channel, to examine the web address of any login page before typing credentials, and to enable multi-factor authentication, which can blunt the value of a stolen password. They also caution that familiarity with a sender's name is no guarantee of authenticity, particularly in fields where reputations are public and easily imitated.

Parker did not respond to a request for comment from Al Jazeera. The report has not prompted any immediate public statement from US authorities about the specific activity described.

The episode nonetheless adds to a growing body of evidence that the contest over AI policy is being fought not only in public forums and bilateral talks but also in the quieter, less visible realm of inboxes and credential theft.