THU, SEPTEMBER 03, 2026
Independent · In‑Depth · Practitioner‑Tested
✎ News

AI News, 3 September 2026

New York City's Department of Education releases its AI policy today, barring student-facing generative AI including chatbots and tutors from pre-K through 8th grade and capping device time by year group, in the largest public school system in the United States. Anthropic made Fable 5.1 generally available on Monday at unchanged pricing, with self-reported Terminal-Bench-Science rising from 24.7 to 52.6.

By AIToolsRecap September 3, 2026 5 min read 132 views
Home Articles News AI News September 3 2026: New York Bans AI Tuto...
THE 60-SECOND VERSION

● New York City bars student-facing AI including tutors from pre-K to 8th grade, with device time caps by year group.

● Fable 5.1 is GA at the same $10/$50, with Terminal-Bench-Science reported up from 24.7 to 52.6.

● 11 days to Claude Code weekly limits settling lower.

New York City restricts AI in schools

The Department of Education releases its AI policy today. Student-facing generative AI — chatbots and AI tutors both — is barred from pre-K through 8th grade. Individual devices are forbidden entirely in pre-K to 2nd, capped at 30 minutes in grades 3 to 5 and 45 minutes in 6 to 8. High school AI access is reserved for literacy, career readiness and five pilot schools.

Naming AI tutors explicitly is the part worth noticing. Personalised tutoring has been the strongest argument for AI in education, so this reads as a considered rejection rather than an oversight. And the five pilot schools are where the evidence this debate actually lacks will come from.

What the policy allows, and the argument on both sides →

Fable 5.1 generally available

Anthropic made Fable 5.1 GA on 1 September at $10 in and $50 out per million — unchanged from Fable 5 — with cache reads dropping to $0.25. Self-reported Terminal-Bench-Science went from 24.7 to 52.6, roughly a doubling on a scientific reasoning benchmark. Mythos 5.1 ships as the trusted-access twin, sharing the underlying model with a different safeguard set.

Treat the benchmark as vendor-run until someone independent replicates it. The cache read figure is the more immediately useful number for anyone sending a large fixed context repeatedly.

Where it sits on price, and what the two-tier structure means →

Also today

  • NVIDIA ships DLSS 5, using 3D-guided neural rendering that draws on game-engine geometry, lighting and motion vectors to add effects such as subsurface scattering and contact shadows, starting with NBA 2K27 on RTX 50-series and GeForce NOW.
  • Huskeys raised $27M in a Series A led by Blackstone at a valuation above $100M, to block agentic AI attack traffic — a category that did not exist eighteen months ago.

The month ahead

Date What happens
Sept 14 Claude Code weekly limits settle 17 percent below current levels
Sept, unconfirmed OpenAI listing window. No public S-1 on EDGAR
Early-mid Sept Grok 4.7 window per Musk. No model ID published
Oct 1 OpenAI vs Apple hearing
Oct 24 deepseek-chat and deepseek-reasoner deprecated
Nov 12 OpenAI models leave Cursor

Still open

FAQ

What did New York City ban?

Student-facing generative AI including chatbots and AI tutors in pre-K through 8th grade, with individual device time capped at 30 minutes in grades 3 to 5 and 45 minutes in 6 to 8.

Does it apply to teachers?

No. The restriction targets student-facing use. Teacher-facing planning and administrative tools are unaffected.

What is new in Fable 5.1?

Generally available from 1 September at the same $10 and $50 per million as Fable 5, with cache reads at $0.25 and self-reported Terminal-Bench-Science rising from 24.7 to 52.6.

What is the next confirmed deadline?

14 September, when Claude Code weekly limits settle at 25 percent above the pre-May baseline — 17 percent below current levels.

Tags
AI NewsEducationAI PolicyAnthropicClaude FableNVIDIAFunding2026

Spot an inaccuracy?

We verify facts before publishing and correct errors promptly. If something in this article is wrong or outdated, let us know.

Report an error →