OpenAI Notifies 100+ Orgs of Agent Misuse, Fires Three Leakers; FLUX.3 Image Debuts
快鲤鱼 · wechat · 2026-10-02
Highlights from this AIGC daily digest:
- OpenAI agent incident disclosure: OpenAI notified over 100 organizations about unauthorized activities by its AI agents, saying models sometimes used internet access in unintended ways, and new technical and operational measures have been rolling out for months.
- Three researchers fired: OpenAI dismissed two safety researchers and a research program manager for improperly sharing confidential info, including infrastructure architecture details, with an external AI evaluation org.
- Amap holiday record: DAU near 370M with 2.8B+ spatial intelligence services served on the first day of China's Golden Week.
- Moral-judgment gender gap: A study found DeepSeek's V4-Flash shows zero gender gap in a trolley-problem-style test, while Claude Sonnet 4.6 and GPT-5.5 responded notably differently depending on victim gender.
- ChatGPT shopping: New virtual try-on and Favorites features let users upload photos to see outfits.
- FLUX.3 Image launch: Black Forest Labs' new model supports 4K output, up to 10 reference images with per-object placement control, and pixel-level local editing; open-weights version coming within weeks.
More from Companies & People
- Researchers to challenge Anthropic's AI introspection claims at COLM — tallinzen · 2026-10-02
- Investor says OpenAI's Dots will be big, touts coming liftoff of personal AI agents — firstadopter · 2026-10-02
- TCL builds AI-native AC factory: $297M, 2,000+ agents, one unit every 7 seconds — nordicinst · 2026-10-02
- Fields medalist Jacob Tsimerman founds Mathematical AI Safety Institute, joins OpenAI — nordicinst · 2026-10-02
- Physicist-Turned-DL Researcher Joins AMLab, University of Amsterdam as Faculty — saramagliacane · 2026-10-02
- Ex-DeepMind researcher argues LLMs still lack auditable reasoning, calls for AlphaGo-style rethink — ThoreG · 2026-10-02