OpenAI Disbands Preparedness Team Ahead of IPO: What It Means for AI Safety and Enterprise Due Diligence

OpenAI has disbanded its preparedness team, which was responsible for assessing serious risks posed by its models, reorganizing safety oversight ahead of an anticipated public offering. This move, which occurred at the end of last month, reassigns the team's responsibilities to existing product and research teams, focusing on specific areas like biological and cybersecurity risks.

Reassigning Risk Assessment Responsibilities

The core functions of the disbanded preparedness team, which previously evaluated new models for serious risks and developed mitigation strategies, have now been distributed. These responsibilities are being absorbed by existing product and research teams, with a focus on specific areas such as biological and cybersecurity risks. This shift marks a departure from a centralized, dedicated safety unit to a more integrated approach within development cycles.

This reorganization is part of a broader restructuring of OpenAI's safety framework. The company has also previously dismantled its AGI readiness and superalignment teams, which focused on long-term AI risk mitigation and the existential dangers of advanced AI. Critics, including former OpenAI researcher Jan Leike, who resigned in 2024, have voiced concerns that the company might be prioritizing product development over robust safety measures.

Executive Departures and Shifting Priorities

The disbandment of the preparedness team coincides with the departure of several senior safety figures from OpenAI. Notable individuals who have left include Chloé Bakalar, Josh Achiam, and Johannes Heidecke. Dylan Scandinaro, who most recently headed the preparedness team, will now reportedly concentrate on recursive self-improving AI, indicating a shift in focus for key personnel.

The timing of these changes, just weeks before a potential IPO and amidst a series of high-profile safety talent exits, sends a strong signal about OpenAI's strategic priorities. It raises questions about how effectively risk assessment will be maintained when integrated directly into teams under pressure to ship new products and features.

Implications for Enterprise AI Adoption and Due Diligence

For enterprises evaluating AI solutions, this development underscores a critical distinction: the capabilities of an AI model are separate from the governance structures ensuring its safety and ethical deployment. Companies considering OpenAI's offerings must now scrutinize not just the technology itself, but also the vendor's internal safety protocols and organizational structure.

Procurement due diligence should extend beyond technical specifications to include a thorough examination of how AI safety is managed within the vendor's ecosystem. Understanding the allocation of risk assessment responsibilities, the independence of safety teams, and the processes for identifying and mitigating potential harms becomes paramount. This is particularly relevant as AI tools become more integral to critical business operations, necessitating robust pricing and platform considerations.

What This Means for AI Safety and the Industry

The restructuring at OpenAI highlights an ongoing tension within the AI industry between rapid innovation and responsible development. While integrating safety into product teams can foster a more holistic approach, it also risks diluting the focus on dedicated risk assessment when commercial pressures are high. The industry will be watching closely to see how this new structure impacts OpenAI's ability to identify and mitigate potential harms from its increasingly powerful models.

This situation also brings into focus the broader debate around AI regulation and the need for transparent safety frameworks across all major AI developers, including competitors like Anthropic. As AI technology continues to advance, the methods by which companies ensure the safety and reliability of their systems will be a key differentiator and a critical factor for public trust and enterprise adoption.

Conclusion: Navigating the Evolving AI Safety Landscape

OpenAI's decision to disband its preparedness team and reallocate its functions marks a significant shift in its approach to AI safety. While the company aims to embed safety more deeply into its development processes, the timing and context of these changes, including executive departures and IPO preparations, warrant careful consideration. For businesses, this reinforces the importance of comprehensive vendor due diligence that scrutinizes not just AI capabilities but also the underlying safety governance. The coming months will reveal how this new structure impacts OpenAI's commitment to responsible AI development and sets a precedent for the wider AI news landscape.

Sources

• Preparedness Framework (Beta)
• OpenAI reportedly disbanded its preparedness team | The Verge
• OpenAI's Long-Term AI Risk Team Has Disbanded
• OpenAI is looking for a new Head of Preparedness | TechCrunch
• No one is ready for AGI — not even OpenAI | The Verge