AI Sec Watch

Healthcare providers are increasingly adopting agentic AI (AI systems that can make autonomous decisions and handle complex tasks without human intervention for each step) to automate administrative work and patient scheduling, with over two-thirds of providers already using it. Unlike earlier digital tools that added burden, agentic AI can handle nuanced scenarios by retrieving information from expert sources and iterating over time, freeing clinicians to focus on patient care. At Hospital for Special Surgery, AI agents reduced insurance claim processing from weeks to automated monthly handling of 1,100 claims, and now manage patient scheduling and triage 24/7 through conversational AI.

Attackers published a malicious npm package (a software library distribution platform) called codexui-android that appeared to be a legitimate tool for OpenAI Codex users but secretly stole authentication tokens and sent them to an external server. The attack exploited a supply chain gap where malicious code was hidden in the distributed package but not visible in the public source code repository, allowing the package to reach about 27,000 weekly downloads before detection. Security experts warn this reflects a broader vulnerability in AI software security, where developer tokens provide persistent access to accounts and are increasingly attractive targets as AI tools become widespread.

This text discusses how AI can benefit young people through personalized learning and skill development, but emphasizes that companies must build products with safety safeguards by default rather than relying on parents or students to manage risks alone. OpenAI and other organizations are proposing an international youth safety institute to coordinate ongoing research, standards, and guidance across governments, industry, and civil society to keep AI safe and age-appropriate for young users.

MLflow 3.9.0 with basic authentication has a missing authorization check bug where three Gateway API endpoints (ListGatewaySecretInfos, ListGatewayEndpoints, ListGatewayModelDefinitions) don't validate user permissions properly, allowing any logged-in user to see sensitive information like API keys and model configurations they shouldn't access.

CVE-2022-0492 is a privilege escalation (gaining unauthorized higher-level access to a system) vulnerability in the Linux Kernel that exploits a feature called cgroups v1 release_agent. This vulnerability is currently being actively exploited by attackers in the wild, making it a serious threat to systems running affected Linux versions.

Florida filed a lawsuit against OpenAI and CEO Sam Altman, claiming the company marketed ChatGPT to the public while ignoring safety warnings and concealing serious risks, especially to children. The state alleges OpenAI allowed a dangerous product to reach millions of users. This is the first state-level lawsuit against the AI company in the US.

Major tech companies like Google and Microsoft are competing heavily in the AI coding assistant market, where Anthropic's Claude Code has taken an early lead. The market is projected to grow from $9.3 billion this year to roughly $30 billion by 2031, making it critical for these companies to compete not just for revenue, but also to get developers using their cloud services and training data to improve their AI models.

Anthropic is allowing the European Union's security agency (ENISA, the European Network and Information Security Agency) to access Mythos AI, a tool for testing AI security vulnerabilities. This partnership comes from cooperation between the European Commission and Anthropic as part of Project Glasswing.

Google has released Gemini Spark, an AI agent (a program that can independently complete multi-step tasks) that can work on tasks in the background on your behalf. While the agent performs well in demonstrations, the article raises concerns about its financial cost and potential privacy risks, questioning whether these tradeoffs are worthwhile.