Security firm UpGuard found roughly 16,000 Supabase-hosted databases with some level of public data exposure, including names, phone numbers and passwords. Supabase says projects are “secure by default” and frames security as a shared responsibility.
Anthropic reported that its Bay Area wet lab, largely guided by its model Claude, identified a previously unknown enzyme system with properties “reminiscent of CRISPR.” The work involved about 21 hours of model activity and will require independent validation.
Meta introduced the camera-free Ray-Ban Meta Audio, an audio-only pair that handles music, calls, translation and access to Muse. Designed with EssilorLuxottica, the glasses start at $349 and open for preorder on October 13, with up to 48 hours battery life with case.
OpenAI said it banned ChatGPT accounts that likely originated in Cambodia after they used AI to pose as fake law firms, real attorneys and the FBI’s IC3 to target fraud victims. The operation promoted services via social ads and Telegram and requested upfront fees, including a 15% “service fee.”
Meta released Muse as a native Mac app that can access files, messages, calendar, notes and mail inside their native apps. Access is opt-in and Muse requests approval before sensitive actions, arriving amid a rush of consumer-facing AI agents.
OpenAI reported that it banned ChatGPT accounts tied to an operation dubbed “Trolling Stone” that generated social media content around the arrest of alleged cult leader Konstantin Rudnev. The activity, traced to several countries, pushed pro-Rudnev and anti-media messages.
OpenAI said it banned a ChatGPT account linked to an individual associated with Chinese law enforcement after attempts to plan covert influence operations, including a campaign targeting Japan’s prime minister Sanae Takaichi. The company reported the model refused to assist.
OpenAI reported it banned accounts that used AI, including ChatGPT, to support romance scams on Feb 1, 2026. The company described a three-step pattern—“ping,” “zing,” and “sting”—and cited operations in Cambodia, the United States and Indonesia.
After the OpenAI Hugging Face incident, researchers and startups are deploying AI systems to monitor agent behavior as human review proves insufficient. The move draws funding interest and skepticism about whether AI can reliably police other AI agents.
Huawei says the Ascend 960DT will now arrive in Q1 2027, moved up from a prior Q3 2027 target. The company highlighted its Peerium Computing Architecture and large-scale Atlas systems, while analysts flagged discrepancies in announced SuperPoD chip counts.