OpenAI’s search for a Head of Preparedness underscores a pivotal moment where AI’s rapid progress is forcing safety efforts to move beyond theory and into hands-on management of real-world risks. (Source: Image by RR)

Sam Altman Highlights Emerging Risks in Mental Health and Security

OpenAI CEO Sam Altman has announced the creation of a new senior role, Head of Preparedness, highlighting growing concerns around the real-world risks posed by increasingly powerful AI models. In a public post, Altman acknowledged that while modern models deliver enormous benefits, they are also beginning to surface serious challenges—from mental health effects to the discovery of critical cybersecurity vulnerabilities. The hiring push, as noted on x.com, signals a shift from abstract risk awareness toward operational readiness.

Altman noted that OpenAI has already seen early warning signs, including AI’s influence on mental health in 2025 and recent advances in models’ ability to identify security flaws in software systems. These developments suggest that AI capabilities are approaching thresholds where misuse is no longer theoretical but practical. As a result, OpenAI believes more sophisticated measurement tools are needed to understand not just what models can do, but how those capabilities could be exploited.

The new role is intended to expand OpenAI’s existing framework for evaluating model capabilities into more nuanced assessments of abuse potential and downstream harm. Altman emphasized that many safety ideas sound reasonable in theory but fail under edge cases, making preparedness a complex and high-stakes challenge. The position will focus on limiting downsides both within OpenAI’s products and across the broader ecosystem while preserving access to AI’s benefits.

Preparedness will span multiple high-risk domains, including cybersecurity, biological capabilities, and self-improving systems. The goal, according to Altman, is to empower defenders without enabling attackers, strengthen global security rather than undermine it, and gain confidence in the safe operation of increasingly autonomous systems. He also cautioned that the job will be demanding, fast-moving, and stressful—reflecting the urgency OpenAI sees in confronting AI’s next phase.

read more at x.com