OpenAI Trial Testimony Reveals How Safety Was Sidelined as the Company Became a Product Business

[encrypted]

A federal court in Oakland heard striking testimony on Thursday from a former OpenAI employee that goes to the heart of what Elon Musk's lawsuit against the company is actually about. Rosie Campbell, who joined OpenAI's AGI readiness team in 2021 and left in 2024 after her team was disbanded, described a company that underwent a fundamental cultural transformation during her time there. "When I joined, it was very research-focused and common for people to talk about AGI and safety issues," she testified. "Over time it became more like a product-focused organization." Her departure coincided with the shutdown of the Super Alignment team, another safety-focused group, in the same period — a pattern that critics argue reflects a deliberate deprioritization of safety work as commercial pressures intensified. Campbell also pointed to a specific incident where Microsoft deployed a version of GPT-4 in India through its Bing search engine before the model had been evaluated by OpenAI's Deployment Safety Board, a procedural bypass she described as dangerous not because of the immediate risk posed by that particular model, but because of the precedent it set for how safety processes would be followed as the technology becomes significantly more powerful.
The testimony adds important texture to the broader legal question at the center of the Musk case: whether OpenAI's transformation from nonprofit research organization to one of the largest private companies in the world broke the founding agreement its early members believed they had made.

The GPT-4 India deployment was notably one of the incidents that contributed to OpenAI's non-profit board briefly firing CEO Sam Altman in 2023, after employees including then-chief scientist Ilya Sutskever and then-CTO Mira Murati raised concerns about his management style. OpenAI's attorneys attempted to blunt the impact of Campbell's testimony by having her acknowledge that OpenAI's safety approach is still, in her opinion, superior to that at Elon Musk's own AI company xAI. But that concession does not address the structural argument Campbell was making: that the company's internal safety processes have become less reliable as commercial priorities have grown, and that unreliable safety processes at a frontier AI lab represent a systemic risk that grows with every increase in model capability.

[AIcash]

they drew ELON like he is some meme coin lol