TECHNOLOGY · AI · CYBERSECURITY · THE SYSTEMS SHAPING BUSINESS
A China-aligned espionage group impersonated a former senior White House official, a prominent economist and a senior Anthropic employee in phishing campaigns aimed at AI policy experts at U.S. universities, think tanks and law firms, The Register reported Thursday. Security firm Proofpoint attributed the activity to a group it tracks as TA419.
What Happened
According to Proofpoint's research, as described by The Register and CyberScoop, the attackers posed as Lynne Parker, a former principal deputy director of the White House Office of Science and Technology Policy, and Heidi Crebo-Rediker, an economist and foreign policy specialist. In February, the group spoofed a senior Anthropic employee in an email to a think-tank AI policy analyst that asked for feedback on the military integration of Claude, Anthropic's AI model.
Starting July 8, the lures shifted to invitations to join a fake AI policy advisory committee or to contribute to a Senate foreign relations report on AI export controls and semiconductor supply chains. The opening messages were harmless and meant to build rapport. Later emails carried links that led to a fake OneDrive loading page, then a Cloudflare Turnstile check, and finally an adversary-in-the-middle phishing page designed to capture Microsoft 365 usernames, passwords and session cookies.
RelatedThe Register reported that the phishing kit combined a modified version of Frameless BitB, an open-source tool that draws a fake browser window inside a web page, with Evilginx phishlets, which relay login traffic so attackers can steal live sessions. Help Net Security said the group typically uses file-sharing and cloud-service themes, relies on Cloudflare's content delivery network and registers domains through NameSilo.
Proofpoint did not identify victims or say whether any accounts were compromised, CyberScoop reported.
Why It Matters
Session theft is the important detail. Because adversary-in-the-middle kits capture the authenticated session, not just a password, they can let attackers into an account even when the target uses standard multifactor authentication. For people who handle drafts of policy papers, congressional input or client work on export controls, a hijacked Microsoft 365 account can expose email, shared documents and contact networks.
The choice of targets reflects how central AI has become to U.S.-China competition. Help Net Security quoted Proofpoint's assessment that the activity likely supports wider Chinese intelligence goals of understanding U.S. AI policy and regulation, amid disputes over export controls and accusations of model distillation. Posing as an Anthropic employee asking about military uses of Claude, and as organizers of a Senate report on chip supply chains, put the lures squarely in that debate.
TA419's Track Record
CyberScoop reported that Proofpoint has documented TA419 activity since April 2025, and The Record said the group has previously targeted U.S. and Japanese think tanks, defense contractors and universities. It has registered look-alike domains, including spellings that mimic the Heritage Foundation and the Japan-Taiwan Exchange Association, as well as one imitating the site of Japanese Defense Minister Shinjiro Koizumi, according to Help Net Security.
The Record paired the report with separate research from Cisco Talos and Symantec on a campaign using a backdoor called Antino, which hit about 350 endpoints across 16 organizations in eight Asian countries between September 2025 and July 2026. The Record presented the two as separate operations and did not attribute the Antino activity to TA419.
What to Watch
Key open questions are whether any of the targeted experts' accounts were breached and whether the group keeps tying lures to upcoming AI policy events, committees and reports. Organizations that work on AI export controls may also review whether their Microsoft 365 setups resist session-cookie theft, for example through phishing-resistant sign-in methods.
