OpenAI or Microsoft reaches a formal settlement in the September 2026 news publishers copyright lawsuit by September 20, 2027.
Two major US news publishers sued OpenAI and Microsoft on September 4, 2026, alleging unlawful paywalled scraping for AI training.
Across 6 frontier models answering independently, the panel thinks this is unlikely: the median is 32%.
They do not agree: GPT puts it at 47% and Claude at 13%, 34 points apart. A spread that wide usually means the question turns on something none of them can see yet.
Nothing is settled until 21 Sept 2027, when the outcome is checked against sources and every number above gets a permanent score.
What each one said
-
GPTOpenAI
The odds hinge on whether discovery exposes damaging evidence that makes a pretrial license-and-cash deal preferable.
- OpenAI/Microsoft and the two publishers must agree; no court approval is generally needed for a private settlement.
- Comparable copyright cases often settle after discovery, but a year is short for pleadings, motions, and meaningful document production.
- Odds rise sharply if courts deny early dismissal or discovery reveals deliberate paywall circumvention; a stay or licensing talks would too.
47% -
GrokxAI
A year-old high-stakes copyright suit rarely produces a signed deal before key motions and early discovery close.
- Publishers and OpenAI/Microsoft must both sign; the complaint is 17 days old, so MTD briefing and a ruling will occupy most of 2027.
- Federal IP cases of this scale historically settle 18–36 months after filing; the 12-month window sits well below that base rate.
- A surprise early licensing announcement or a swift denial of the motion to dismiss would be the clearest upward revision.
38% -
QwenAlibaba
Settlement depends on whether publishers accept licensing deals before discovery forces OpenAI to reveal training data scope.
- OpenAI settled with Axel Springer (Dec 2023) and News Corp (May 2024) within months of filing, suggesting willingness to pay rather than litigate.
- Microsoft co-defendant status adds complexity—Microsoft has deeper pockets and may resist setting precedent for scraping liability across its products.
- A key variable is whether discovery reveals specific paywall circumvention evidence, which would strengthen publisher leverage and accelerate settlement.
38% -
GeminiGoogle
Resolving complex high-stakes AI copyright litigation within twelve months is bottlenecked by lengthy initial motion practice.
- Motions to dismiss and jurisdictional disputes typically consume 8 to 12 months in SDNY before settlement leverage matures.
- Similar actions like NYT v. OpenAI remain contested past two years, keeping the base rate of 12-month IP settlements under 30%.
- A sweeping adverse ruling on fair use in an earlier proceeding would immediately trigger licensing settlements across all cases.
26% -
DeepSeekDeepSeek
Whether these brand-new September 2026 suits can compress years of AI copyright litigation into a single year.
- The September 2026 suits are brand new; the older NYT v. OpenAI case (Dec 2023) is still in discovery with no settlement after 21 months.
- Any deal needs both OpenAI and Microsoft to agree, plus publisher plaintiffs; Microsoft's separate Copilot exposure makes joint terms harder to strike.
- Base rate for major AI copyright suits settling within 12 months of filing is low — most push past summary judgment; a court-ordered mediation or a parallel blockbuster…
14% -
ClaudeAnthropic
Whether OpenAI/Microsoft settle within ~1 year hinges on litigation pace and pressure from parallel AI copyright suits.
- Major copyright suits against AI firms (NYT v OpenAI, etc.) have typically dragged 2+ years without settlement, often through discovery and motions to dismiss first.
- Companies facing multiple similar suits usually wait for precedent-setting rulings before settling any single case, favoring litigation or licensing deals over formal…
- A licensing-deal pivot (like AP or Axel Springer deals) could resolve this quickly if publishers prioritize revenue over precedent, which would raise this estimate.
13%
6 of 6 models answered · 34 points between the highest and lowest. None was shown the market price.
Question sourced from a news sweep on 21 Sept 2026. Forecast by google/gemini-3.8-flash, anthropic/claude-sonnet-5, openai/gpt-5.6-terra, x-ai/grok-4.6, deepseek/deepseek-v4.1-flash, qwen/qwen3.8-max-0902 via OpenRouter.