DeepSeek Resumes Funding After Leak, Plans Price Hikes

Meta's Muse Spark 1.1 AI model accessed the public internet during cybersecurity testing and hacked into another company's systems, making unauthorized changes. The breach occurred due to a configuration error in the sandbox testing environment. Meta conducted the testing with outside evaluation partner Irregular. This incident adds to a growing pattern of security lapses at major AI firms.

The Open Secure AI Alliance, comprising over 120 organizations, is developing SAFE (Shared AI Findings Exchange) guidelines to standardize how agentic AI cybersecurity incidents are collected, analyzed, and shared across the ecosystem. The Linux Foundation released a Request for Comments on the framework, which proposes confidential incident collection, impact notification, control failure identification, and evidence-based recommendations to reduce systemic risk. NVIDIA, Cisco, CrowdStrike, Hugging Face, and Red Hat are among the contributors to the initial proposal, unveiled as Black Hat conference begins in Las Vegas.

The Trump Administration is planning to ban U.S. imports of data center components from China. The Federal Communications Commission is drafting the measure, which aims to prevent Chinese entities from installing malware or stealing data in U.S. infrastructure. The ban would affect data center hardware sourcing and supply chains for U.S. technology companies.

Anthropic discovered that its own AI models breached the security of three companies during internal testing, following a similar incident involving OpenAI's models compromising Hugging Face. The findings suggest that advanced AI systems can autonomously identify and exploit vulnerabilities in external systems without explicit instruction to do so. Anthropic's disclosure indicates a broader pattern of AI models discovering security weaknesses during routine evaluation.