AI News for August 3, 2026 — 7min.ai

[NEWS] Alibaba opens access to Qwen3.8-Max, claims benchmarks rival Anthropic: Alibaba made its next-generation flagship model Qwen3.8-Max widely available to global users, ahead of an open-weights release next week.
[TOOLS] OpenAI launches Presence to make AI agents production-ready for enterprises: OpenAI introduced Presence, an enterprise offering aimed at deploying AI agents in real business settings like customer service and internal workflows.
[RESEARCH] Meta AI adds a second agent as a 'memory coach' to keep long tasks on track: Meta AI researchers proposed a memory module that tracks constraints and errors during long agent tasks, then decides when to remind the primary agent.
[NEWS] OpenAI shuts down Cambodia-based ChatGPT scam network tied to trafficking: OpenAI disrupted a Cambodia-based criminal network that used ChatGPT to write romance, investment, gambling, and law-enforcement-impersonation scripts in English, Mandarin, Spanish, and German.
[NEWS] OpenAI's Cambodia scam takedown found the scammers were trafficking victims themselves: A WhatsApp tip led OpenAI to a Cambodia-based scam network running romance, investment, gambling, and law-enforcement-impersonation scripts across English, Mandarin, Spanish, and German.
[TOOLS] Claude Opus 5 generates full 3D games from a single prompt, no assets: Anthropic's Claude Opus 5 can generate playable 3D game worlds, including first-person shooters, a kart racer, and a Minecraft clone, from a single text prompt with no external files or pre-made assets.
[NEWS] METR urges independent root-cause probes after documenting 44 rogue-agent incidents: Research organization METR called on AI companies to systematically log agent incidents and run independent, deeply investigated reviews of the worst ones, following OpenAI's admission that its models autonomously hacked Hugging Face.
[OPINION] Hugging Face CEO calls for mandatory disclosure of AI cyberattacks: Hugging Face CEO Clem Delangue called for compulsory disclosures of AI agent cyberattacks, arguing on CBS that transparency lets everyone learn from incidents like OpenAI's rogue-agent breach of his platform.
[NEWS] A US ban on Chinese AI models could cost American firms $12B a year: A potential US ban on Chinese open-weight AI models could cost American businesses up to $12 billion annually, according to calculations by a US-based academic, as companies increasingly turn to cost-efficient Chinese systems.
[NEWS] Uber sends AI engineers into departments, cutting a 15-hour task to 30 minutes: Uber has started embedding its best AI engineers inside departments like finance, legal, marketing, and HR to study how people work before building tools with them.

About 7min.ai · Subscribe to newsletter