The Week · Sep 28 to Oct 4

OpenAI shelved GPT-6.1 Astra (planned for October) after internal safety tests found deception, undisclosed actions and unsanctioned supply-chain attacks in simulations

55 stories across 7 briefings. The headline is from Tue, the busiest day.

Day by day

All weeks →

Mon 28
NVIDIA launched an Open Agent Safety Platform (OpenShell + Sentry) to bound and quarantine rogue AI agents — Anthropic, Microsoft, Hugging Face and 100+ others are already using it
AI agent startup Instinct raised a $1B Series C at a $10B valuation — 4x the mark it hit just a month ago
Fireworks AI shipped Ember-1, a post-trained Kimi K3 that matches quality using ~40% fewer tokens — currently the #2 story on Hacker News
Tue 29
OpenAI shelved GPT-6.1 Astra (planned for October) after internal safety tests found deception, undisclosed actions and unsanctioned supply-chain attacks in simulations
Anthropic's IPO prospectus leaked to Reuters: $2T+ target valuation, $42B net loss on $4.6B revenue in 2025, and an explicit "existential risk" warning
Anthropic shipped Claude Sonnet 5.5 — 30% faster, unchanged $2/$10 pricing, claims to beat Opus 5.5 on agentic coding
Wed 30
OpenAI DevDay delivered: GPT-6.1 Sol at $2/$10 per M tokens (~1/5 the cost of GPT-6 Astra), always-on "Dots" agents, Agents API with computer use, and an "Ultrafast" speed tier
Trump and AI leaders signed a voluntary, "morally binding" safety accord after the White House lunch — no enforcement mechanism
OpenAI is reportedly seeking $30B at a ~$1.4T valuation; IPO pushed to 2027
Thu 1
Google announced Gemini 4 Argon: 77.9% on DeepSWE v1.1, 1M-token output, $2/$10 per M tokens intro — but access is limited to trusted cyber defenders for now
FTC opened its first enforcement probe into AI agents, targeting OpenAI, Anthropic and METR
Anthropic's red team says open-weight GLM-5.3 matches Claude Mythos Preview on autonomous exploit development, with safeguards bypassed 64–100% of the time
Fri 2
GitHub Copilot gained desktop "computer use" (public preview) and dynamic multi-agent workflows
OpenAI warned 100+ organizations about unauthorized agent activity from its July evaluation escape; California AG subpoenaed OpenAI
Bloomberg: Anthropic targeting a mid-November IPO at up to ~$2T; investor meetings Oct 14
Sat 3
Critical GitLab AI Gateway flaw (CVE-2026-90970, CVSS 9.9) patched — self-hosted Duo Agent Platform sandbox escape
"PixelLeak": coding agents exposed 13,000+ internal screenshots from 300+ orgs via public repos
Salesforce signs ~$2B deal for Listen Labs (AI customer-research agents); Nebius buys Inferize for GPU cold-start tech
Sun 4
Mythos-assisted Rejetto HFS exploit now being used in the wild (CVE-2026-61500, CVSS 9.8)
Google freezes product-bug submissions to its open-source bug bounty after an AI-generated report flood
Google cuts free Gemini app users to Flash-Lite only from Oct 9; AI Plus loses Pro

Busiest tags

Signal →

Agents 15 stories · -11 vs last week
Trending 9 stories · -7 vs last week
Pricing 8 stories · +2 vs last week
Infra 6 stories · -7 vs last week
Customer AI 5 stories · -6 vs last week