AI Sec Watch

This document proposes a strategy for the U.S. government to create lasting institutions that oversee frontier AI (the most advanced AI systems being developed). The plan has three main parts: build a national framework based on state laws already in place, strengthen CAISI (the federal organization responsible for frontier AI safety) as the main federal institution, and develop a broader government-wide plan to address national security and public safety risks from advanced AI.

Google is rolling out a new Android security feature called 'fake call detection' that protects users from AI deepfake scam calls where scammers impersonate someone's contacts. The feature works by having a user's device send an encrypted confirmation signal when receiving a call, and if that signal is missing, it pings the actual contact's phone to verify the call is real, warning the user to hang up if the contact's device confirms they're not calling.

The UK's Competition and Markets Authority has ruled that Google must allow website publishers to opt out of AI Search features, including AI Overviews (summaries generated by AI) and prevent their content from being used to train Google's AI models. This new rule gives publishers, especially news organizations, more control over how their content is used by AI systems.

Debapt is a system that uses multiple AI agents (independent AI programs that work together) debating with each other to build profiles of APT (advanced persistent threat, a sophisticated type of cyberattack) adversaries by analyzing cyber threat intelligence (information about security threats). The system uses an ontology (a formal structure that defines how concepts relate to each other) to organize this debate process. This research proposes a new way to understand and track advanced attackers by having AI agents discuss and reason through threat data together.

Anthropic expanded its Project Glasswing (an AI-based vulnerability hunting initiative that finds security bugs in software) to 150 more companies, especially those in critical infrastructure like power and healthcare. However, security experts warn this creates a bottleneck problem: if AI finds vulnerabilities 10 or more times faster than before, companies may not be able to validate, prioritize, patch, and deploy fixes quickly enough, potentially overwhelming security teams rather than actually improving defense.

This research paper presents a method for detecting deepfakes (synthetic videos or images created by AI to look realistic) using a dual-mode Swin Transformer, which is a type of neural network architecture. The approach uses multi-scale feature learning (analyzing visual details at different zoom levels) and local ambiguity mitigation (reducing confusion in uncertain areas) to improve detection accuracy. This is a technical contribution to security research, not a response to an existing vulnerability or security incident.

Palo Alto Networks CEO Nikesh Arora reported a surge in customer meetings, with the company fielding roughly 1,200 inquiries in recent weeks from organizations seeking guidance on AI security risks. The article notes that AI-powered attacks are becoming more sophisticated, making cybersecurity more important for companies, and that earlier investor concerns about AI disrupting cybersecurity companies appear to have been overblown.

LibreChat (a ChatGPT-like tool that connects to multiple AI providers) has a security flaw in versions up to 0.8.3 where someone with editing access to a shared agent can delete files globally, breaking the owner's separate private agents that use the same files. This is a cross-agent integrity violation, meaning one agent's access should not affect another agent's files.

LibreChat, a ChatGPT-like application supporting multiple AI providers, has a vulnerability in versions up to 0.8.3 where users with limited VIEW access can retrieve encrypted admin passwords and API keys through specific API endpoints, exposing credentials that should remain secret. This happens because the API returns plaintext sensitive values instead of hiding them from non-admin users.