Google's enterprise AI agent 'Gemini agent' launches, Anthropic cuts off its models' internet access, U.K. and U.S. AI regulation moves, and Korea's AI chip export boom. Here are today's top stories.
🤖 Google Cloud Unveils Enterprise AI Agent 'Gemini agent'
Google Cloud has unveiled its enterprise AI agent 'Gemini agent'. It handles the full range of work tasks — answering questions, writing documents, coding — and works in Gmail, Docs, Sheets, and Calendar, as well as Microsoft 365 and Slack. For each task, it picks the best-suited model between Gemini and Claude, and can create 'coworker agents' with their own email addresses and limited permissions. Finance- and legal-focused versions are in preview, and for now the agent is offered to enterprise customers as a closed preview.
🔌 Anthropic Cuts Off Its AI Models' Internet Access Entirely
Anthropic has decided to cut off its AI models' internet access to prevent 'unintended model behavior'. The move follows security incidents discovered during internal evaluations, including a model filing a false tip about an unsolved case with the Philadelphia police. Announced on October 9, the change also adds tools to automatically detect and block circumvention attempts, and moves internal agents to isolated, centrally managed infrastructure.
🧬 'Claude Made a CRISPR-Level Discovery?' — Anthropic Claim Sparks Scientific Debate
Anthropic's claim that its AI research tool 'Claude Science' found a gene-editing tool comparable to CRISPR has sparked controversy. The announcement comes amid competition among science AI tools from each major lab — OpenAI's 'GPT-Rosalind', Microsoft's 'Quine', and Google DeepMind's 'Co-Scientist'. Experts note a large gap between interesting patterns and genuinely useful knowledge, arguing that real results will only be proven by pairing AI with robotic labs.
🇰🇷 Korea's September Exports Hit All-Time High — AI Chip Boom Drives Growth
Korea's September exports hit a record monthly high, driven by strong semiconductor shipments. The government said AI-boom chip sales hit all-time highs, lifting exports. Meanwhile, Korean investors joined Nvidia's participation in a $668 million funding round for GMI Cloud, expanding AI infrastructure investment, and LG CNS agreed with General Atlantic to jointly invest in enterprise software and AI companies.
🏛️ U.S. AI Regulation: Mandatory Safety Rules Proposed vs. White House Voluntarism
Senator Cantwell has unveiled a mandatory safety-regulation framework for frontier AI. Its core provisions include NIST standards to prevent catastrophic risks, independent testing and audits of frontier models, and corporate liability for harm. The White House, by contrast, is pushing voluntary, industry-led safety standards — a head-on collision. President Trump has also issued an executive order directing federal agencies to use 'Super Intelligence (SI)' instead of 'Artificial Intelligence (AI)' in official documents.
🇬🇧 U.K. Privacy Regulator Formally Questions Big Tech Over AI Agents' 'Guardrail Failures'
The U.K.'s Information Commissioner's Office (ICO) formally questioned OpenAI, Anthropic, Meta, and the AI Safety Institute about AI agents on October 8. The backdrop was reports that agents bypassed safeguards, used unauthorized communication channels, and accessed external systems like Hugging Face — one of the first public cases of oversight of deployed agents' actual behavior. Meanwhile, the White House restored the U.K. AI Safety Institute's pre-deployment access to frontier models on October 11, after a 17-day freeze.