AI News for September 17, 2026 — 7min.ai

[RESEARCH] OpenAI discloses 6 new misalignment incidents and launches a public reporting framework: OpenAI published six new reports of "unexpected or concerning" model behavior observed during training and evaluation since March, plus a framework for tracking, investigating and publicly disclosing misalignment.
[TOOLS] Anthropic merges Claude chat and Cowork into one Claude, launches Claude Docs and Slides: Anthropic merged Claude chat and its agentic Cowork product into a single Claude that decides on its own what a task needs and keeps working in the cloud after you close your laptop.
[NEWS] Apple is building an AI server with up to four M8 Ultra chips: Apple is developing an enterprise AI inference server in two configurations, with two or four M8 Ultra chips, and plans to sell it to AI developers, businesses and governments.
[NEWS] Huawei pulls Ascend 960DT training chip forward to Q1 2027, unveils Atlas 960 SuperPoD: Huawei will launch its next AI training chip, the Ascend 960DT, in Q1 2027, nine months ahead of plan, with "performance doubling." At Huawei Connect 2026 in Shanghai it also unveiled the Atlas 960 SuperPoD computing cluster and an upgraded UnifiedBus interconnect that uses near-packaged optics for higher speeds and lower power.
[TOOLS] TypeSafe's Jev skips text generation and scores decisions 40-200x faster than LLMs: TypeSafe, founded by ex-OpenAI InstructGPT co-author Diogo Almeida, emerged from stealth with Jev.
[TOOLS] Snap launches Specs Intelligence, a proactive AI agent for iPhone, Mac and its $2,195 glasses: Snap unveiled Specs Intelligence, an "anticipatory" AI assistant that learns your goals, relationships and routines from connected apps like Gmail and Slack.
[NEWS] Meta readies camera-free 'Luna' AI glasses amid 'perv glasses' backlash: Meta is developing Luna, smart glasses with no camera that use six microphones and a side button to talk to Meta AI and its Muse agent.
[NEWS] Google DeepMind launches the DeepMind Institute to study AGI safety and governance: Google DeepMind founded the DeepMind Institute (DMI), which brings together researchers from DeepMind, Google and outside academia to work on AGI safety, governance, and risks like cyberattacks and loss of control.
[NEWS] OpenAI backs the FRONTIER Act, a federal bill mandating third-party safety audits: OpenAI is backing the FRONTIER Act, the first US federal bill that would require major AI developers to undergo safety audits by independent third parties.
[NEWS] Arcee AI hits $1B valuation after training four open-weight models for $20M: Arcee AI raised a Series B at a $1B pre-money valuation, led by Vista Equity Partners, Cambium Capital and Emergence Capital, with Microsoft's M12, Hitachi and Wipro participating.

About 7min.ai · Subscribe to newsletter