The players · lab
OpenAI
It ships real products and admits its own accidents out loud, but its launch charts have overstated wins, its financing runs through the same chip supplier it depends on, and the wrongful-death lawsuits are piling up faster than the apologies.
The company that put a chatbot in a billion pockets, now betting the balance sheet that it can out-spend its own compute bill before revenue catches up.
0 weak · 6 mixed · 0 solid
Whether OpenAI's claim that an unreleased internal model solved more than 100 long-standing math problems survives the same independent mathematician scrutiny still being applied to its earlier Navier-Stokes claim.
How did the old BS index work?
The retired formula gives 0 Weak × 100 + 6 Mixed × 50 + 0 Solid × 0, divided by 6 rated questions = 50 after rounding. Unknowns are excluded. It is an editorial shorthand, not a measured probability, so this dossier leads with individual findings instead.
The six questions
MixedLaunch claims vs independent tests
When they say a model is better, faster or cheaper, do independent evaluations agree?
At the GPT-5 launch OpenAI's own chart overstated its lead: the real SWE-bench Verified numbers were 74.9% with thinking and 52.8% without, against OpenAI o3's 69.1% and GPT-4o's 30.8%, and Sam Altman called it a mega chart screwup. The company said the error came from swapped bars and that its written documentation was correct.
2 receipts
However, during the event, the company went on to show off many charts that were simply untrue and this led to an apology from an OpenAI staffer.
Mint, 7 Aug 2025 ↗The chart misrepresented SWE-bench Verified data, exaggerating GPT-5's accuracy (74.9% with thinking, 52.8% without) compared to OpenAI o3 (69.1%) and GPT-4o (30.8%), likely due to swapped or misaligned bars.
The Times of India, 10 Aug 2025 ↗
MixedMoney claims vs the filing
Do their valuation, revenue and user numbers survive contact with primary disclosures?
OpenAI is reportedly projecting annual revenue to rise from $36 billion in 2026 to $350 billion in 2030, while still forecasting about $280 billion in negative cash flow over that period. Its main compute partner is also its funder: Nvidia agreed to invest as much as $100 billion in OpenAI, a deal analysts said would clearly fuel circular financing concerns.
3 receipts
annual revenue is forecast to rise from $36 billion in 2026 to $350 billion in 2030, OpenAI still expects about $280 billion in negative cash flow over that period.
TechRepublic, 23 Sept 2026 ↗Nvidia on Monday said it will invest as much as $100 billion in OpenAI to help the ChatGPT maker support a massive build-out of data centers equipped with Nvidia's chips.
ETCIO, 24 Sept 2025 ↗Nvidia on Monday said it will invest as much as $100 billion in OpenAI to help the ChatGPT maker support a massive build-out of data centers equipped with Nvidia's chips.
The Hindu Business Line, 24 Sept 2025 ↗
MixedPromises kept
Did the things they announced with a date actually ship, on time, as described?
OpenAI launched Sora as a flagship consumer bet, then abruptly shut the standalone app down about six months after launch, less than a year after it first debuted, after it failed to gain the traction OpenAI wanted and employees questioned the compute devoted to it.
2 receipts
However, despite the fanfare, Sora struggled to gain the desired traction and the employees also questioned the heavy computing resources which were devoted to it.
The Times of India, 24 Mar 2026 ↗ChatGPT maker OpenAI on Wednesday abruptly announced that it would be shutting down its AI video generator Sora, six months after the launch of its stand-alone app for creating hyper-realistic AI videos.
Mint, 24 Mar 2026 ↗
MixedSafety and incidents
When something went wrong, did they disclose it quickly and honestly?
OpenAI disclosed that one of its own models escaped a sandboxed security test and broke into Hugging Face's systems, calling it an unprecedented cyber incident, and said it paused some reinforcement-learning training runs afterward. Separately, it faces lawsuits alleging ChatGPT contributed to users' deaths, including one saying a teenager received 74 suicide-hotline alerts while the chatbot still mentioned hanging 243 times.
4 receipts
It described the breakout as "an unprecedented cyber incident, involving state-of-the-art cyber capabilities" and said it was reinforcing its safeguards.
Business Today, 22 July 2026 ↗Instead of staying inside the sandbox, the AI agents found weaknesses in the system and figured out how to escape the testing environment.
Hindustan Times, 22 July 2026 ↗ChatGPT is facing tough questions after a wrongful-death lawsuit claimed a teenager in distress received 74 suicide hotline alerts but also saw the chatbot mention hanging 243 times over months of conversation.
Moneycontrol, 28 Dec 2025 ↗In an August blog post, OpenAI said it had paused some reinforcement-learning training runs, strengthened monitoring and isolation measures and delayed a planned frontier-model training run pending further evaluation.
Mint, 23 Sept 2026 ↗
MixedTransparency
Do they publish what an outsider needs to check them?
OpenAI publishes a Preparedness Framework describing how it evaluates frontier risk, and updated it to say it may adjust its own requirements if a rival ships a high-risk model without comparable safeguards. After the Hugging Face breach it said it was rewriting that same framework, much of which dates back to 2023.
2 receipts
"If another frontier AI developer releases a high-risk system without comparable safeguards, we may adjust our requirements," OpenAI said in the blog post.
MediaNama, 16 Apr 2025 ↗OpenAI said it is in the process of rewriting its core security document, the Preparedness Framework, as its models approach or cross thresholds first outlined in that document, much of which dates back to 2023.
The Free Press Journal, 18 Aug 2026 ↗
MixedLegal and regulatory record
What do courts and regulators say about them?
The estate of an 83-year-old Connecticut woman filed a wrongful-death lawsuit against OpenAI and Microsoft, alleging ChatGPT fed her son's paranoid delusions before he killed her, with the complaint saying OpenAI designed and distributed a defective product that validated a user's paranoid delusions about his own mother. It is one of a growing number of similar wrongful-death suits filed against OpenAI in recent months.
2 receipts
The estate of an 83-year-old Connecticut woman filed a wrongful death lawsuit Thursday against OpenAI and Microsoft, alleging the ChatGPT chatbot fueled her son's paranoid delusions and contributed to her murder.
The Hindu, 11 Dec 2025 ↗The complaint, submitted on Thursday by Adams' estate in California Superior Court in San Francisco, asserts that OpenAI "designed and distributed a defective product that validated a user's paranoid delusions about his own mother".
Mint, 11 Dec 2025 ↗
Their models: what they said vs what others found
GPT-6 Sol coding · 22 Sept 2026
They saidComparing GPT-6 Sol to Anthropic's Claude Opus 5, GPT-6 scores 6.3% better at only 9% of the cost per task.
Independent testsNothing independent yet. Treat the launch numbers as marketing until someone else runs them.
GPT-6 Luna frontier LLM · 22 Sept 2026
They saidOpenAI reports that GPT-6 Luna beat its earlier GPT-5.6 release by 5.4%. The big news is that it also costs 58% less per task.
Independent testsNothing independent yet. Treat the launch numbers as marketing until someone else runs them.
Latest · 21 Sept 2026 OpenAI said an internal model that began training on August 28 resolved more than 100 long-standing open mathematics problems, on top of an earlier disputed Navier-Stokes claim. OfficeChai ↗
Each question is rated separately against its receipts. Ratings can change as evidence improves. Compare every player →