THE DAILY BRIEF

The US Treasury Department's technology team is seeking access to Anthropic's Mythos AI model to hunt for vulnerabilities. This comes amid broader government scrutiny of the company, including DOD supply chain risk designation.

► Government agencies are now conducting red-team operations on proprietary AI systems, establishing precedent for state-level security audits that could shape AI governance and vendor requirements.

Adobe released patches for Acrobat DC, Reader DC, and Acrobat 2024 to fix a zero-day vulnerability that hackers actively exploited for at least four months. The flaw affected widely-used document reading software.

► Multi-month exploitation of flagship consumer software demonstrates that even major vendors struggle to detect active zero-days, leaving millions of users exposed to data theft and malware.

An attacker purchased 30 WordPress plugins outright and implanted backdoors into all of them, distributing malicious code to thousands of WordPress sites through the official plugin ecosystem.

► Open-source supply chain risks escalated from dependency injection to outright acquisition of popular projects, making plugin vetting impossible for site owners without deep code audits.

Microsoft is exploring deployment of always-on AI agents within Microsoft 365, including OpenClaw technology, to automate enterprise workflows. The company is testing autonomous systems for worker productivity.

► Shift from chat-based AI to persistent autonomous agents in core workplace software signals AI is moving from user-initiated tools to background decision-makers, raising workplace surveillance and job displacement concerns.