OpenAI wins Musk suit, faces backlash

OpenAI won the case Elon Musk brought over its shift toward a for-profit structure, but the courtroom record did not leave the company with a clean reset. A federal jury in Oakland rejected Musk’s claims on May 18, handing Sam Altman and OpenAI a legal victory that Reuters said removed the biggest immediate threat to a future public offering. Yet the same trial aired testimony from former colleagues who described Altman as dishonest, putting his credibility into the public record just as OpenAI pushes deeper into business and consumer finance. A separate pressure point opened in California days earlier. TechTimes reported that a proposed federal class action alleges OpenAI routed private ChatGPT conversations to third parties through tracking tools, and that the suit was filed just before OpenAI launched bank-account, portfolio and credit-card connections for ChatGPT Pro users. (money.usnews.com) Datadog’s latest benchmark added another complication for OpenAI’s pitch around automation. ARFBench, a benchmark built from 63 real production incidents, found GPT-5 was the top-performing model tested at 62.7% accuracy, but still behind human domain experts at 72.7%. ### Why did winning Musk’s lawsuit still leave OpenAI exposed? Reuters reported that Musk’s lawsuit sought to show OpenAI had improperly abandoned its nonprofit mission, but the case also became a venue for former insiders to challenge Altman’s honesty. (techtimes.com) Eight witnesses cited by Musk’s lawyer said Altman had misled or lied to others, according to Reuters’ trial account. (arxiv.org) Sam Altman answered that charge directly on the stand. “I believe I am an honest and trustworthy businessperson,” he testified, according to Reuters. James Rubinowitz, a trial lawyer and AI specialist quoted by Reuters, said the verdict removed the largest legal threat to an IPO, but said the trial record left “the worst documentary evidence about its governance” in public view. (money.usnews.com) ### What exactly does the California privacy suit allege? TechTimes reported that the proposed class action in California accuses OpenAI of secretly routing users’ private ChatGPT conversations to third parties. According to that report and follow-on coverage surfaced in search results, the complaint alleges ChatGPT’s web interface used Meta Pixel and Google Analytics in ways that transmitted sensitive user data without proper consent. (money.usnews.com) May 15 is central to the timing. TechTimes and TechCrunch reported that OpenAI launched a personal-finance feature for ChatGPT Pro users in the United States on that date, allowing connections to bank accounts, investment portfolios and credit cards. That rollout tied a privacy dispute to a product aimed at more sensitive financial workflows. (techtimes.com) ### What did Datadog’s benchmark actually test? Datadog said ARFBench was built from real internal incidents rather than synthetic examples. The company and the paper’s authors said the benchmark includes 750 questions across 142 time series and 5.38 million data points drawn from 63 production incidents sourced from Datadog telemetry. GPT-5 led the model field, but the paper said models and human experts showed “complementary strengths.” The researchers reported a model-expert “oracle” could reach 87.2% accuracy, above either humans or models alone, suggesting current systems still work better as support tools than as replacements in incident response. (techtimes.com) That framing comes from the paper and Datadog’s write-up, not from OpenAI. ### Why do these threads converge now? (arxiv.org) OpenAI’s finance push, the California privacy allegations and the post-trial questions around Altman all bear on trust. Reuters said the Musk verdict simplified OpenAI’s path toward an IPO that could value the company at up to $1 trillion, while TechTimes said the privacy case arrived as OpenAI expanded into linked financial accounts. (arxiv.org) The next milestones are concrete. The California class action will proceed through early court filings unless dismissed or settled, and OpenAI’s finance features for ChatGPT Pro users are already live in the United States, where the company is testing whether users will hand over more sensitive data while those allegations remain unresolved. (techtimes.com) (money.usnews.com)