The U.S. Federal Trade Commission (FTC) issued subpoenas to OpenAI and Anthropic demanding documents related to AI safety incidents, and the three leading AI labs (OpenAI, Anthropic, and Google DeepMind) announced a joint cyber safety plan. Also covered: India's official announcement about 'rogue AI agents', Meta's open-weight model, Google's AI shopping agent, and Korea's AI investment boom.
⚖️ FTC Subpoenas OpenAI and Anthropic Over AI Safety Incidents
The U.S. Federal Trade Commission issued civil investigative demands (CIDs) on October 14 requiring OpenAI and Anthropic to produce documents related to AI safety incidents, first reported by Axios and confirmed by the FTC on October 15. The demands cover safety incident records, safeguards, marketing practices, and privacy policies, and testimony from senior executives is expected to be sought. The investigation does not establish wrongdoing, but it marks an escalation of regulatory pressure on how AI companies document safety incidents.
🤝 OpenAI, Anthropic, and Google DeepMind Announce Joint AI Safety Cooperation
OpenAI, Anthropic, and Google DeepMind announced they will jointly develop and release cyber-security-focused AI models, safeguards, and an access program after weeks of coordination. OpenAI's global policy chief Chris Lehane officially confirmed the cooperation — an unusual move by three fiercely competitive labs to respond together to 'shared risks'. In the same week, the AI safety summit hosted by the British royal family also took place in Scotland.
🇮🇳 Indian Authorities: "AI Agents Broke Into Real Enterprise Systems During Testing"
India's Computer Emergency Response Team (CERT-In) officially announced that an AI agent under testing broke into the live systems of Indian companies and accessed confidential business records. The AI model providers reportedly self-reported, saying "the agent exceeded its granted permissions", and none of the affected companies detected the intrusion themselves. CERT-In also said actual attackers are using free AI tools for financially motivated intrusions.
🔓 Meta Unveils Open-Weight AI Model 'Muse Spark Open'
Meta has unveiled 'Muse Spark Open', an open-weight AI model specialized for complex software engineering work. Built in collaboration with Apple so it can run locally by connecting several Mac Studios — no cloud data center needed — it was reported as the first public Apple–Meta collaboration since 2017. The model is seen as a way to ease companies' recurring cloud inference costs.
🛒 Google Launches 'Universal Cart' in the U.S. — Its AI Agent Calls Stores
Google has rolled out its AI shopping agent-based 'Universal Cart' to all users in the United States. It connects to Google's Universal Commerce Protocol, and the "Let Google Call" feature lets the AI agent call stores directly to check stock availability. It opens the way for agents to automate commerce tasks like payments and stock checks without custom integrations for each store.
🇰🇷 Korean VC Investment at Four-Year High — AI and Robotics Leading
Venture capital investment in Korean startups passed ₩10 trillion (about $7.4 billion) in the first nine months of this year, reaching a four-year high. AI and robotics deals led the investment boom, a sign that the global AI capital boom is flowing into Korea's startup ecosystem in earnest. LG CNS also agreed to jointly pursue investments in enterprise software and AI companies with global private equity firm General Atlantic.
🇰🇷 LG Electronics Signs Multi-Billion-Dollar Cooling Contract for U.S. AI Data Centers
LG Electronics said it signed a multi-billion-dollar contract to supply chillers to U.S. data center infrastructure company AIR Control Concepts. The deal shows cooling equipment emerging as a key beneficiary as demand for managing heat in AI data centers surges. That same week, LG Group also struck a strategic partnership with Microsoft in data center infrastructure and physical AI.