The Musk-Altman trial is dominating tech coverage, but the more important OpenAI story this week is quieter. A new arXiv paper by Xiao Wang found that chain-of-thought reasoning models, the kind OpenAI has been heavily promoting, don't reduce bias. They amplify it. The longer the reasoning chain, the more pronounced the position bias becomes. This lands at an awkward moment: Sam Altman has been on the stand defending OpenAI's governance while simultaneously signaling, via TurboFund's tracked investor intelligence, that GPT-5.5's compounded improvements in speed and memory create quote, super-additive user value. More capable and more biased are not mutually exclusive. That's the problem.
Reasoning Length as a Liability
Wang's paper is technically precise about something the AI industry has been vague about. The assumption, embedded in the entire chain-of-thought paradigm, is that more deliberation produces better outputs. The paper calls this the length-driven position bias: models that think longer tend to over-weight early positions in a sequence, regardless of content quality. This matters enormously for legal applications, where AI is already flooding courts with fake citations, and for financial analysis, where Perplexity is launching 35 finance-specific agentic workflows this week. TurboFund's Signal Report shows 6 high-conviction investor signals in AI/LLMs this week alone, nearly all bullish. The bias research is not priced into that sentiment.
Regulatory Capture and the Governance Gap
A separate arXiv paper this week, Big AI's Regulatory Capture by Birhane et al., maps how the AI industry has systematically shaped its own regulatory environment over the past decade. The Musk-Altman trial is a live demonstration of what happens when governance is left to personal agreements between founders rather than structural frameworks. Both men agreed on OpenAI's mission in 2015 and now disagree on what it meant. That interpretive gap is exactly what regulatory capture is designed to exploit. The paper argues that industry interference and government complicity have produced a governance vacuum that no individual lawsuit can fill. The trial is theater. The vacuum is structural.