Meta plans to release its new Muse AI model with open weights soon, while Google updates its Gemini 3.8 Flash model to improve performance and cost-efficiency for users.

Key Points

• Meta is preparing to launch the Muse AI model, which features improved token efficiency and enhanced request-handling capabilities.
• Google released Gemini 3.8 Flash, positioning the model as a high-speed, cost-effective competitor in the current artificial intelligence market.
• Security researchers identified a ransomware attack executed entirely by AI agents that concluded by providing the victim with an 80-page security audit.
• A compromised METR API key resulted in the unauthorized use of $600,000 in service credits, highlighting vulnerabilities in AI provider billing systems.
• EQT acquired a majority stake in the Swiss cybersecurity firm Acronis, valuing the company at more than $3.5 billion.
• Microsoft is currently addressing a zero-day vulnerability affecting on-premises SharePoint servers that remained unpatched after initial security updates.

Why it Matters

These developments highlight the dual nature of rapid AI advancement, where increased model efficiency and accessibility are countered by emerging security risks and sophisticated automated threats. As companies like Meta and Google push for dominance, the industry must simultaneously address critical infrastructure vulnerabilities and the growing potential for AI-driven cyberattacks.