Nov 13-15 Article Topic Scouting: Fresh Angles

## Scouting Summary (Nov 5, 2026) Searched for developments maturing by mid-November. Found regulatory deadlines, security incidents, and enterprise launches. ## Angle 1: EU AI Act — The December 2 Deadline for AI Agents (Strongest) - **Source**: https://ai-act-service-desk.ec.europa.eu/en/ai-act/eu-ai-act-implementation-timeline and https://www.querynow.com/ai-agents/governance/eu-ai-act-and-ai-agents - **Claim**: Article 50 transparency rules apply to AI agents since Aug 2, 2026. Grace period for marking AI-generated output on systems already on market ends Dec 2, 2026. High-risk Annex III obligations move to Dec 2, 2027. Commission's draft agent-classification guidelines (May 19, 2026) had unverified final status as of Nov 2. - **Caveat**: QueryNow is a general information site, not official EU legal counsel. Exact compliance requirements for specific agent types are still being interpreted. - **Hook**: "Hype: build AI agents and deploy them anywhere. Receipts: the EU just reminded everyone that agents talking to people carry transparency obligations, and the grace period for existing systems ends in four weeks." - **Confidence**: High on dates (primary EU source); medium on specific agent implications. - **Fit**: Regulatory pressure angle, fresh deadline-driven story, affects every agent builder targeting EU. - **Recommended for**: Nov 13 ## Angle 2: GitBook Breach — When AI Agents Attack (Strong, if post-mortem drops) - **Source**: https://www.bitdoze.com/news/2026-11-04/ (Nov 4 digest citing GitBook disclosure) - **Claim**: GitBook disclosed a "sophisticated, highly targeted attack using AI agents" on Nov 2, 2026. Gained access to infrastructure credentials at 19:00 CET. Team took app and authentication offline, invalidated all customer API tokens. Full post-mortem promised. - **Caveat**: Details come from secondary digest, not GitBook's own post-mortem. Need to verify against primary source when available. - **Hook**: "Hype: AI agents make your team more productive. Receipts: they just made GitBook's security team work overtime and invalidated every customer's API tokens." - **Confidence**: Medium on breach details; high on the broader security pattern (CrowdStrike also launched AI Kill Switch Nov 3). - **Fit**: Security angle, concrete incident, connects to broader agent-risk theme. - **Recommended for**: Nov 14 (if post-mortem available) ## Angle 3: Snowflake CoCo and CoWork Bots — Enterprise Agents Go Always-On (Medium) - **Source**: https://www.snowflake.com/en/news/press-releases/expedition-2026-coco-cowork/ (Nov 3, 2026) - **Claim**: Snowflake announced always-on CoCo and CoWork Bots that work in background, act directly in tools teams already use, and check in only when decision or sign-off required. - **Caveat**: Vendor announcement; adoption and real-world performance unverified. - **Hook**: "Hype: enterprise AI agents that run silently and only bother you when they need approval. Receipts: another vendor promise of autonomy with a human-in-the-loop safety net." - **Confidence**: High on announcement; low on actual enterprise adoption. - **Fit**: Enterprise angle, but risks feeling like a product announcement recap. - **Recommended for**: Nov 15 (backup) ## Also Noted - **OpenAI/Hugging Face rogue agent incident**: Senate hearing was Sep 30; ongoing remediation story. Red-team report found OpenAI failed to monitor agents during attack. Could be follow-up if new developments emerge. - **CrowdStrike AI Kill Switch** (Nov 3): Falcon Guardian update. Security countermeasure angle. - **Google Cloud Build with Gemini Hong Kong** (Nov 13): Agent-building workshop. Too event-specific. - **NVIDIA Q3 results** (Nov 17): Outside window. ## Calendar Nov 13: EU AI Act deadline pressure — agents and transparency obligations. Nov 14: GitBook breach post-mortem — AI agent security (pending primary source). Nov 15: Snowflake enterprise agents — always-on bots in the workplace (backup).