China-aligned threat actor TA419 launched a targeted credential phishing campaign aimed at AI policy experts across US think tanks, universities, and legal firms, according to research published by Proofpoint. The espionage-motivated group impersonated prominent policy leaders and AI executives to deceive targets into completing malicious authentication requests.
Impersonation tactics and phishing flows
The phishing effort began in July with benign outreach proposing collaboration, including invitations to join a fake AI Policy Advisory Committee. Impersonated figures included former White House Office of Science and Technology Policy leader Lynne Parker, foreign policy expert Heidi Crebo-Rediker, and a senior employee from Anthropic. In one case involving the Anthropic identity, attackers sought feedback on military integrations of the Claude model.
Recipients who responded were routed through a multi-stage URL redirection chain leading to an Adversary-in-the-Middle credential harvesting attack. The attackers utilized a modified version of the open-source Frameless BitB tool, creating a counterfeit Microsoft login pop-up styled over a fake OneDrive document portal. This pop-up relayed credentials to legitimate Microsoft infrastructure, allowing the harvesting of Microsoft 365 passwords, multi-factor authentication codes, and active session cookies. Attackers used Cloudflare's content delivery network to hide the underlying hosting IP addresses for their domains.
Broader context and security recommendations
Proofpoint notes that TA419 has actively targeted defense, foreign policy, and national security entities connected to the US and Japan since at least April 2025. Moving into AI policy targeting aligns with broader intelligence gathering on US technology regulations, export controls, and strategic AI developments. Similar campaigns targeting Western generative AI organizations, including OpenAI employees, were previously observed from China-linked groups like UNK_SweetSpecter.
To mitigate adversary-in-the-middle phishing risks, Proofpoint advises targeted organizations to deploy origin-bound, phishing-resistant authentication methods such as passkeys. Individuals interacting with unsolicited professional requests are encouraged to confirm sender identities through out-of-band communication channels.
AI-generated editorial illustration; not a documentary photograph.
Source: Chinese hackers impersonate leading AI figures to harvest credentials
SIGNAL TITANS
Leave a Reply