AI Daily Brief — 12 February 2025
Wednesday delivered OpenAI’s most explicit product roadmap of the year and the first US court ruling against an AI training fair-use defense. Sam Altman tweeted that GPT-4.5 (internally “Orion”) — the company’s last non-chain-of-thought model — was coming “in weeks,” with GPT-5 “in months” as a unified system merging the o-series reasoning and GPT-series. Free-tier ChatGPT users will get unlimited GPT-5 access at a standard intelligence level; Plus and Pro tiers get higher settings. Goal: end the model picker. Harvey, the legal-AI startup, raised a $300-million Series D at a $3-billion valuation led by Sequoia. Judge Stephanos Bibas’s Feb 11 partial summary judgment against Ross Intelligence — the first US court ruling rejecting fair use for AI training — broke into wide industry coverage. Paris-summit fallout dominated international press.
Top stories
- Sam Altman publishes OpenAI roadmap: GPT-4.5 “in weeks,” GPT-5 “in months,” model picker retired. GPT-4.5 will be OpenAI’s last non-chain-of-thought model. GPT-5 to merge o-series reasoning and GPT-series. Free tier gets unlimited GPT-5 access at standard intelligence; Plus/Pro tiers get higher settings. Goal: make AI “just work.” via Sam Altman on X · via Axios
- Harvey raises $300M Series D at $3B valuation. Sequoia-led, with Coatue, Kleiner Perkins, OpenAI Startup Fund, GV, Conviction, Elad Gil, and REV (LexisNexis parent RELX’s VC arm). Valuation doubled vs the July round. CEO Winston Weinberg said ARR had crossed $50M with $100M tracking within 8 months; customer count grew from 40 to 235 across 42 countries in 2024. via Harvey
- Thomson Reuters v. Ross — first US court rejects AI fair-use defense. Judge Stephanos Bibas (D. Del.) partial summary-judgment ruling from Feb 11 reached wide industry coverage Feb 12: Ross Intelligence’s use of 2,243 Westlaw headnotes to train its legal-research AI was NOT fair use. First US court ruling on the fair-use question for AI training. Sets precedent for NYT/OpenAI, Getty/Stability, and music-publisher cases. via Davis Wright Tremaine
- Paris-summit declaration fallout dominates international press. The US/UK refusal — Vance citing “ideological bias,” UK Downing Street citing missing “practical clarity” — broke into Al Jazeera, China Daily, and global front pages as a sharp split with the Bletchley/Seoul safety consensus. via Al Jazeera
Who shipped
No model launches. OpenAI shipped a roadmap; Harvey shipped a funding round; Thomson Reuters won a ruling. Anthropic, Google DeepMind, Meta, and xAI stayed quiet on the product side.
Open-source pulse
No major Chinese or open-weight release Feb 12. Cohere’s Command A would land March 13; xAI’s Grok 3 was five days away; Anthropic’s Claude 3.7 Sonnet twelve days away. The week’s open-source story was post-Summit absorption of Mistral’s sovereign-EU partnerships and the Current AI foundation funding.
Money, infra & hardware
Harvey’s $300M Series D plus the rolling Paris commitments (€109B France + €200B EU InvestAI + €150B coalition) anchored the day’s investment ledger. The Vance speech reset US policy framing for capital-allocation decisions across hyperscaler boardrooms.
Quiet corners
Modi reiterated India would host the next Summit (Feb 19-20 2026 New Delhi). The AI Foundation and Council for Sustainable AI moved into setup mode. Macron’s office circulated execution plans for the €109B package.
By the numbers
- “weeks” / “months” — GPT-4.5 / GPT-5 ship horizons per Altman
- $300M / $3B — Harvey round size / valuation
- $50M → $100M — Harvey ARR trajectory
- 2,243 Westlaw headnotes at issue in Thomson Reuters v. Ross
- Most-mentioned lab: OpenAI
Compiled by AI Feed’s editor from verified web sources for 12 February 2025.