China-Linked TA419 Targets US AI Policy Experts with Sophisticated Phishing
A China-aligned cyber espionage group, TA419, has targeted US artificial intelligence policy experts with phishing campaigns impersonating government officials and AI company employees to steal cloud account credentials.
A sophisticated phishing operation, attributed to the China-aligned threat actor TA419, has specifically targeted individuals involved in U.S. artificial intelligence policy. Researchers at Proofpoint revealed that the campaign, which began in July, employed social engineering tactics by impersonating prominent figures such as Lynne Parker, former principal deputy director of the White House Office of Science and Technology Policy, and economist Heidi Crebo-Rediker. The initial emails aimed to establish trust by inviting recipients to join advisory committees or contribute to reports on AI export controls and supply chains, rather than immediately soliciting credentials.
Once a target engaged with the initial message, TA419 would send a shortened link, purportedly containing more information. This link would then redirect the victim through a series of websites before presenting a convincing fake Microsoft OneDrive sign-in page. This elaborate setup was designed to capture both user credentials and active browser session information, enabling an adversary-in-the-middle attack. In this type of attack, the victim believes they are interacting with legitimate Microsoft infrastructure, completing multi-factor authentication prompts and passing access checks, all while the attacker harvests the session data.
Proofpoint noted that TA419 utilized a modified version of the open-source phishing tool known as Frameless BitB. This tool is capable of creating a false browser window within a webpage, mimicking familiar sign-in prompts. In this specific operation, the tool was used to overlay a fake Microsoft login window onto a page that appeared to be a legitimate OneDrive document-sharing site, making the deception highly effective.
The group's activities also included a separate campaign in February where TA419 impersonated a senior employee from Anthropic, a leading AI company. In this instance, the phishing email targeted an AI policy analyst at a U.S. think tank, seeking their input on the controversial topic of military applications of Anthropic's Claude AI models. This demonstrates the group's keen interest in sensitive and high-profile areas of AI development and policy.
While Proofpoint did not identify specific victims or confirm account compromises, the research indicates that TA419 has been actively targeting individuals associated with U.S. and Japanese think tanks, defense contractors, universities, and law firms since at least April 2025. The group's focus on AI policy appears to be an extension of its pre-existing interest in defense, national security, energy, international relations, and foreign policy.
TA419 has also demonstrated its operational sophistication by registering domain names that closely resemble those of real organizations, including the Heritage Foundation, the World Economic Forum, and the Japan-Taiwan Exchange Association. This tactic further enhances the credibility of their phishing lures and increases the likelihood of successful compromise.
The report does not directly attribute these cyber espionage activities to the Chinese government, which has consistently denied such operations. However, the targeting of U.S. AI policy experts aligns with broader geopolitical tensions and competition between the U.S. and China over advancements in artificial intelligence, export controls, and national security implications.
Proofpoint has made indicators of compromise available on their website for security professionals to detect and defend against TA419's ongoing operations. The campaign highlights the growing threat of state-sponsored cyber espionage focused on critical emerging technologies like artificial intelligence.
This latest report from Infosecurity Magazine elaborates on the TA419 campaign, specifying that the threat actor impersonated AI policymakers and economists. The group successfully compromised Microsoft 365 accounts through these spearphishing efforts, underscoring the increasing trend of adversaries leveraging AI-related themes for social engineering tactics.
This latest report details the sophisticated technical methods employed by TA419, including the use of a modified Frameless BitB toolkit and a Browser-in-the-Browser technique. The attackers leveraged fake OneDrive and Microsoft login pages, combined with a custom JavaScript module, to capture session cookies and MFA codes, enabling them to bypass standard credential theft defenses and gain persistent access to cloud accounts.
This new reporting details the specific tactics used by TA419 in their July 2026 phishing campaigns, including the impersonation of a former White House official and the use of fake advisory committees and Senate reports to lure targets. The article further elaborates on the technical execution, highlighting the use of attacker-in-the-middle (AitM) techniques via custom domains and a Browser-in-the-Browser (BitB) overlay to steal Microsoft 365 credentials, and mentions specific domains used in these attacks.