← Back to scans

Will a court find that OpenAI has infringed the copyright of the New York Times? — New York Times wins

NYTOAI-27DEC31 · Science and Technology · 2026-04-05
14%
Multi-Step
63%
Market Price
-49.0%
Edge
62%
Confidence
Final Rationale
The NYT v. OpenAI case has no trial date set as of early 2026 with only ~21 months until the Jan 2028 deadline, making a dispositive ruling unlikely (~22%). Even conditional on reaching a ruling, NYT's win probability on memorization claims is ~55% (strong for near-verbatim reproduction but novel fair use questions remain). The Devil's Advocate correctly identified that I underweighted partial summary judgment paths, NYT's strategic incentive to litigate rather than settle, and the strength of memorization claims — pushing my estimate from 7% to 14%. I diverge sharply from the 63% market because the procedural posture (no trial date, active discovery disputes, MDL consolidation of 16 cases) makes a pre-2028 liability finding extremely unlikely; the low-volume market likely prices 'NYT eventually wins' rather than the specific SDNY-finding-before-2028 condition.
What Would Change My Mind
A trial date being set for 2027, or summary judgment motions being filed on memorization claims in 2026, would push this above 25%. If the court issued a partial ruling on any count (e.g., denying fair use defense on regurgitation claims at summary judgment), that would push toward 40%+. Conversely, a settlement announcement would drop this to near 0%.
1. Planning Agent Sonnet
binary legal kalshi_data court_docket article_search web_search wikipedia_lookup code_execution
Sub-questions
IDQuestionWeight
sq1 Will the NYT v. OpenAI case proceed to trial (rather than settle or be dismissed) before the resolution date of Jan 1, 2028? 35%
sq2 If the case proceeds to trial or summary judgment, will the Southern District of New York find OpenAI liable on at least one count before Jan 1, 2028? 35%
sq3 Will the case reach a dispositive ruling (verdict or summary judgment) before the Jan 1, 2028 deadline? 30%
Reasoning Approach
This question has two distinct hurdles: (1) procedural — will the case reach a dispositive ruling before Jan 1, 2028, and (2) substantive — will the NYT prevail on at least one count. The key causal factors are: (a) case timeline — filed Dec 2023, resolution deadline Jan 2028, leaving ~4 years total but only ~2 years remaining from today; (b) settlement probability — high-profile IP cases involving AI have strong incentives to settle (OpenAI faces existential risk, NYT wants precedent and damages); (c) trial pace — complex copyright cases in SDNY typically take 3-5+ years; (d) merits — AI training data copyright is genuinely novel law, courts could go either way on fair use. The critical distinction from superficially similar cases: this is NOT asking if NYT eventually wins, but if SDNY specifically finds liability before Jan 1, 2028. The tight timeline makes a pre-deadline verdict highly unlikely even if the case is strong. Discovery disputes, motion practice, and the novelty of AI copyright law all create delay. Base rate: most major copyright cases settle (~70%) and the few that go to trial take years. The AI copyright legal landscape is still being established (Andersen v. Stability AI, Authors Guild cases) and courts are moving cautiously. A settlement before 2028 would resolve NO. The probability of a YES resolution requires the case to (1) not settle, (2) reach a liability finding, and (3) do so before Jan 2028 — each step substantially reducing probability.
2. Research Execution No LLM
ToolStatusTime
kalshi_data OK 1.91s
court_docket OK 0.54s
article_search OK 0.74s
web_search OK 50.33s
wikipedia_lookup OK 2.36s
code_execution OK 0.0s
3. Evidence Synthesis Sonnet · 17 items
#StrengthDirectionSourceClaimPriced In
1 STRONG ↑ UP web_search Judge Sidney Stein denied OpenAI's motion to dismiss in April 2025, allowing main copyright infringement claims to proceed to trial, but no trial date has been set as of March 2026. Yes
2 STRONG ↓ DOWN web_search As of March 2026, the case is still in discovery/pretrial proceedings; evidence gathering including executive depositions is ongoing, indicating trial is not imminent. Yes
3 MODERATE ↓ DOWN web_search The NYT case is part of consolidated MDL pretrial proceedings for 16 copyright lawsuits against OpenAI, adding complexity and likely extending timeline. Yes
4 MODERATE ↓ DOWN web_search OpenAI counter-alleged that the NYT secretly deleted evidence of its internal use of OpenAI models — discovery disputes are active and unresolved, further delaying trial. No
5 MODERATE ↓ DOWN web_search Experts state no court will decide fair use in AI training until summer 2026 at the earliest, suggesting the NYT case trial itself is further out. Yes
6 WEAK ↓ DOWN code_execution A base rate calculation estimates ~55% probability of settlement before the deadline, ~40% chance of proceeding to a ruling, and only ~35% chance that ruling occurs before Jan 1, 2028 — yielding low compound probability. No
7 MODERATE ↓ DOWN article_search OpenAI is preparing for a major IPO and raised $110B+ in funding in early 2026, suggesting financial resources to sustain litigation and potentially strong incentives to settle to remove legal uncertainty before going public. Yes
8 STRONG ↑ UP web_search The court denied OpenAI's motion to dismiss direct infringement and contributory infringement claims, meaning the substantive copyright claims have survived initial legal challenge and appear facially strong. Yes
9 STRONG — NEUTRAL web_search OpenAI's primary defense is fair use for AI training data — legal doctrine that is novel, unsettled, and has not yet been adjudicated by any court as of early 2026, creating genuine uncertainty on the merits. Yes
10 MODERATE ↓ DOWN article_search The Supreme Court ruled unanimously in March 2026 in Cox Communications v. Sony Music that ISPs cannot be held liable for subscriber infringement, narrowing secondary liability theories — potentially relevant to contributory infringement claims. No
11 MODERATE — NEUTRAL wikipedia_lookup AI copyright law in the US remains deeply unsettled as of 2023-2026, with multiple pending lawsuits and no definitive court ruling on whether AI training on copyrighted data constitutes fair use. Yes
12 MODERATE ↑ UP web_search NYT's case is strengthened by evidence that ChatGPT can reproduce NYT articles near-verbatim (memorization/regurgitation claims), which may be harder for OpenAI to defend under fair use than pure training data ingestion. Yes
13 STRONG ↓ DOWN web_search No trial date has been set as of March 2026; case is in discovery phase with active disputes, making a dispositive ruling before Jan 1, 2028 — only ~21 months away — highly unlikely given SDNY complex case timelines. Yes
14 MODERATE ↓ DOWN web_search Fair use in AI training is not expected to be decided by any court until summer 2026 at earliest (in other cases), and NYT v. OpenAI appears further behind those cases procedurally. Yes
15 WEAK ↓ DOWN code_execution Statistical model estimates only ~35% probability that if the case goes to ruling, that ruling arrives before Jan 1, 2028, based on typical SDNY complex copyright case timelines. No
16 MODERATE ↓ DOWN web_search MDL consolidation with 16 AI copyright cases adds procedural complexity that typically extends rather than accelerates individual case timelines. Yes
17 MODERATE ↑ UP kalshi_data The Kalshi market for 'New York Times wins' is currently priced at 63%, up 8% over the past 30 days, with average daily volume of 108 contracts — a substantial divergence from analytical base rates. Yes
Information Gaps
  • No data on current settlement negotiation status or whether any formal settlement talks are underway between NYT and OpenAI/Microsoft
  • No specific trial scheduling order or case management conference dates from the SDNY docket
  • No information on the current discovery schedule — key depositions completed, pending, or disputed
  • No data on comparable AI copyright case timelines (e.g., Authors Guild v. OpenAI, Andersen v. Stability AI) that could inform base rates for this specific legal context
  • No information on whether NYT or OpenAI has filed or plans to file motions for summary judgment and on what schedule
  • No analysis of whether OpenAI's anticipated IPO creates settlement incentives or litigation risk disclosure obligations that might accelerate resolution
  • No data on Judge Stein's historical case management pace or his scheduling of complex IP cases
  • Missing: any 2026 reporting specifically on NYT v. OpenAI discovery rulings or scheduling orders
Key Uncertainties
  • Whether OpenAI's anticipated IPO creates strong incentives to settle before going public, which would resolve NO
  • Whether any of the 16 MDL AI copyright cases will produce a fair use ruling that effectively predetermines NYT's case outcome
  • The pace of discovery — active disputes over evidence destruction allegations could significantly delay or accelerate the case
  • Whether Congress might pass AI copyright legislation that moots or alters the litigation landscape before 2028
  • The Cox Communications Supreme Court ruling's impact on contributory infringement claims in the NYT case
  • Whether the case proceeds on its own track or remains fully consolidated with the MDL proceedings, which affects scheduling control
  • OpenAI's financial trajectory — if fundraising/IPO proceeds well, OpenAI has resources to fight; if struggling, settlement incentives increase
Market Context
Price: 0.63 · 7d: +0.03 · 30d: +0.08
Volume: LOW-MODERATE — 108 contracts/day is thin for a market with significant uncertainty · Depth: UNKNOWN — not provided in data
Reliability: LOW — The 63% market price appears to dramatically overstate the probability of a pre-Jan 2028 liability finding given the case's current procedural posture; market may be pricing 'NYT eventually wins' rather than the specific condition of SDNY finding liability before Jan 1, 2028; low volume suggests limited price discovery; possible market misinterpretation of resolution criteria
4. Independent Forecaster Opus · No market price Did NOT see market price
Independent estimate: 7% (72% confidence)
sq1: Will the NYT v. OpenAI case proceed to trial (rather than settle or be dismissed) before the resolution date of Jan 1, 2028?
25%
Base rate: ~30% of major copyright cases go to trial rather than settling. Evidence pushing DOWN: (1) Case is still in discovery as of March 2026 with active disputes, no trial date set — this strongly suggests trial is unlikely within 21 months. (2) MDL consolidation with 16 cases adds complexity and delay. (3) OpenAI's IPO preparations create strong settlement incentives. (4) Discovery disputes (evidence destruction allegations) add further delay. Evidence pushing UP: (1) Judge denied motion to dismiss, so case hasn't been thrown out. (2) NYT may want precedent rather than settlement money, reducing settlement probability somewhat. Net: The combination of no trial date, active discovery disputes, MDL complexity, and OpenAI IPO incentives makes proceeding to trial before Jan 2028 unlikely. I estimate ~25%.
sq2: If the case proceeds to trial or summary judgment, will the SDNY find OpenAI liable on at least one count before Jan 1, 2028?
45%
This is conditional on the case reaching a dispositive ruling. Base rate for plaintiff success in copyright infringement cases that reach trial: roughly 50-60%. Evidence pushing UP: (1) Claims survived motion to dismiss, suggesting facial strength. (2) Memorization/regurgitation evidence (near-verbatim reproduction) is strong on direct infringement. Evidence pushing DOWN: (1) Fair use doctrine is genuinely unsettled for AI training — courts could go either way. (2) Cox Communications ruling may weaken contributory infringement theories. (3) Novel legal territory means courts may be cautious. Evidence NEUTRAL: No court has yet ruled on fair use for AI training, so there's genuine 50/50 uncertainty on the core legal question. The memorization claims are somewhat stronger than pure training-data claims, pushing slightly above 50%. But the novelty and unsettled nature of the law creates real downside risk. I estimate 45% conditional on reaching a ruling.
sq3: Will the case reach a dispositive ruling (verdict or summary judgment) before the Jan 1, 2028 deadline?
15%
Base rate: Complex copyright cases in SDNY typically take 3-5+ years from filing to trial. Case filed Dec 2023, so by Jan 2028 that's ~4 years — at the very fast end of the range. But as of April 2026, no trial date is set and the case is still in discovery with active disputes. That leaves only ~21 months. Evidence pushing DOWN: (1) No trial date set — in SDNY, trial dates are typically set 6-12 months in advance, so even if set tomorrow, trial might not occur until late 2027 at earliest. (2) MDL consolidation adds delay. (3) Discovery disputes are unresolved. (4) Fair use rulings in other cases not expected until summer 2026 at earliest. (5) Summary judgment motions would need to be briefed, which typically takes months after discovery closes. Evidence pushing UP: (1) Summary judgment is possible without a full trial, and could theoretically be filed on some counts. (2) Partial summary judgment on memorization/regurgitation claims could happen faster. Net: Even accounting for the possibility of partial summary judgment, the timeline is extremely tight. I estimate 15%.
Show full reasoning trace
base rate: Complex copyright cases in SDNY take 3-5+ years. Case filed Dec 2023 with ~4 years to deadline. ~70% of major copyright cases settle. No court has yet ruled on AI training fair use. Plaintiff win rate at trial in copyright cases is roughly 50-60%.
evidence updates: Key evidence pushing probability DOWN: No trial date set as of March 2026 with only 21 months remaining (strong negative signal). MDL consolidation with 16 cases adds complexity. Active discovery disputes. OpenAI IPO incentives favor settlement. Key evidence pushing UP: Motion to dismiss denied on main claims. Memorization evidence is strong. NYT may prefer precedent to settlement. Net: The timeline constraint is the dominant factor — even if NYT has a strong case on the merits, reaching a liability finding in SDNY before Jan 2028 is very unlikely given current case posture.
combination method: Conditional: P(ruling before deadline) × P(NYT wins | ruling) = 0.15 × 0.45 ≈ 0.07. Settlement probability (~55-60%) further reduces the chance of any ruling occurring.
final: The overwhelming constraint is the timeline. With no trial date set, active discovery disputes, MDL complexity, and strong settlement incentives (especially around OpenAI's IPO), the probability of SDNY finding OpenAI liable before January 1, 2028 is very low — approximately 7%. The merits question (whether NYT would win if it went to trial) is genuinely uncertain (~45%), but the procedural/timeline hurdle is the binding constraint.
5. Devil's Advocate Sonnet Market price revealed
Reasoning Flaw Asymmetric Evidence Missing Info
Challenges
  1. The forecaster may be interpreting the resolution criteria too narrowly. The question asks 'Will a court find that OpenAI has infringed the copyright of the New York Times?' — this could potentially resolve via summary judgment on specific counts (e.g., memorization/regurgitation claims) without a full trial. Partial summary judgment on direct infringement for near-verbatim reproduction could move faster than the full fair use determination. The forecaster acknowledges this possibility but may underweight it.
  2. The 15% estimate for reaching a dispositive ruling by Jan 2028 may be too low. The case was filed Dec 2023, and by Jan 2028 that's 4 full years. Discovery disputes don't necessarily mean the case is stalled — they can indicate active, progressing litigation. Many complex IP cases in SDNY do reach summary judgment within 3-4 years. The forecaster notes 3-5+ years as typical but then treats 4 years as 'very fast end of range' — it's actually mid-range. Summary judgment motions could be filed in late 2026 or early 2027 and decided by late 2027.
  3. The forecaster's settlement probability (~55-60%) may be too high given NYT's apparent strategic objectives. The NYT has publicly framed this as an existential issue for journalism and copyright protection. Major media organizations rarely settle landmark cases when they believe they can set favorable precedent. The NYT has the financial resources to litigate fully and strong institutional incentives to establish case law. OpenAI's IPO incentives cut both ways — OpenAI might want certainty, but NYT might prefer to maintain litigation pressure.
  4. The conditional probability of NYT winning given a ruling (45%) may be too low. The memorization/regurgitation evidence is unusually strong for a copyright case — near-verbatim reproduction is textbook infringement, and fair use is much harder to establish for outputs that closely reproduce protected expression. Even if the broader 'training is fair use' question is 50/50, the specific memorization claims are substantially stronger. A partial ruling on direct infringement for reproduced content could be well above 50%.
  5. The forecaster dismisses the market price as likely misinterpreting resolution criteria, but offers no evidence that market participants are confused. The market has been trading for some time with 108 contracts/day — not high volume but enough for basic price discovery. The 8% upward trend over 30 days could reflect informed participants tracking case developments. While low-volume markets can be mispriced, dismissing a 63% price requires stronger justification than speculation about misinterpretation.
  6. The MDL consolidation may not delay the NYT case as much as assumed. MDL proceedings often have lead cases or bellwether trials, and the NYT case — as the highest-profile and earliest-filed — could be the lead case. MDL consolidation is for pretrial proceedings only; the case could be remanded for trial or resolved via summary judgment within the MDL. Some MDL proceedings actually accelerate resolution of key legal issues by consolidating briefing.
Suggested adjustment: +7pp
The forecaster systematically underweights several factors: (1) the possibility of partial summary judgment on memorization claims moving faster than a full trial, (2) the strength of NYT's case on direct infringement for near-verbatim reproduction, (3) NYT's strategic incentive to litigate rather than settle, and (4) the possibility that MDL proceedings could accelerate rather than delay key rulings. Adjusting P(dispositive ruling) from 15% to ~22% and P(NYT wins | ruling) from 45% to ~55% would yield ~12%. Combined with modest upward adjustment for partial summary judgment scenarios, a total probability of ~14% is more appropriate. The +0.07 adjustment reflects these reasoning corrections without anchoring to the market price.
Market comparison: Forecaster: 7%, Market: 63%. This is an enormous divergence. The forecaster's reasoning about timeline constraints is strong and well-supported — the case's procedural posture does make a pre-2028 finding unlikely. However, the forecaster may be somewhat too pessimistic on timeline (15% for dispositive ruling) and on NYT's conditional win probability (45%). The market at 63% likely does reflect some misunderstanding of resolution criteria or is pricing the broader 'NYT eventually wins' sentiment, but the forecaster's complete dismissal of the market signal is aggressive. A modest upward adjustment to ~14% is warranted based on reasoning corrections, not market anchoring.
6. Final Forecast Opus
14% (62% confidence)
The NYT v. OpenAI case has no trial date set as of early 2026 with only ~21 months until the Jan 2028 deadline, making a dispositive ruling unlikely (~22%). Even conditional on reaching a ruling, NYT's win probability on memorization claims is ~55% (strong for near-verbatim reproduction but novel fair use questions remain). The Devil's Advocate correctly identified that I underweighted partial summary judgment paths, NYT's strategic incentive to litigate rather than settle, and the strength of memorization claims — pushing my estimate from 7% to 14%. I diverge sharply from the 63% market because the procedural posture (no trial date, active discovery disputes, MDL consolidation of 16 cases) makes a pre-2028 liability finding extremely unlikely; the low-volume market likely prices 'NYT eventually wins' rather than the specific SDNY-finding-before-2028 condition.
7. Entry Decision Opus Trader
Side
NO
Position Size
$1500
Contracts
4054
Trader Confidence
62%
Trade Rationale
The forecaster identifies a massive 49% edge — market prices NYT winning at 63% while the forecast is only 14%. The core reasoning is compelling: no trial date is set, there are active discovery disputes, MDL consolidation of 16 cases creates delays, and only ~21 months remain until the Jan 2028 deadline. The market appears to be pricing the eventual outcome rather than the specific condition of a court finding before 2028. Even after the Devil's Advocate pushed the estimate up from 7% to 14%, this remains an enormous edge. The sub-question decomposition is rigorous — only 15% chance of reaching a dispositive ruling before deadline, and even conditional on that, only ~45% NYT wins. Forecaster confidence at 0.62 is moderate but the edge is so large it warrants a substantial position.
Allocation Logic
Despite the enormous edge (~49%), I'm sizing at $1500 rather than $2000 because: (1) liquidity is low-moderate at 108 contracts/day, (2) the 635-day horizon introduces significant uncertainty about procedural developments, and (3) the forecaster confidence is moderate at 0.62. Still above baseline given the exceptional edge magnitude.
Entry price: $0.37
Current: $0.40
Status: OPEN
P&L: $121.62
Pipeline Timing
StepTime
Planning 29.42s
Research 50.33s
Synthesis 45.84s
Forecaster 31.49s
Critic 26.86s
Final 23.44s