TL;DR
OpenAI has opened a search for a new Head of Preparedness to lead its effort on emerging AI risks across areas such as cybersecurity and mental health. The role ties into the company’s Preparedness Framework; the hire follows earlier reorganizations and recent updates to OpenAI’s safety approach.
What happened
OpenAI put out a job listing for a Head of Preparedness charged with carrying out its Preparedness Framework — the company’s approach to identifying and preparing for frontier AI capabilities that could cause severe harm. CEO Sam Altman flagged several areas of concern in a social post, noting models are beginning to surface challenges from mental-health impacts to capabilities that find critical computer-security vulnerabilities. Altman urged candidates who can balance equipping defenders with advanced tools while preventing attacker misuse to apply, and he highlighted related responsibilities such as handling biological capabilities and assessing systems that can self-improve. The search comes after OpenAI established a preparedness team in 2023 to track risks ranging from phishing to speculative threats like nuclear misuse. In the past year, OpenAI reassigned Aleksander Madry — previously Head of Preparedness — to an AI reasoning role, and other safety leaders have shifted roles or left. The company also revised its Preparedness Framework recently and said it might alter safety requirements if rival labs release high-risk models without comparable protections.
Why it matters
- Leadership in preparedness affects how OpenAI anticipates and mitigates risks from increasingly capable models.
- Capabilities that can locate security flaws or influence mental health create cross-disciplinary safety and regulatory challenges.
- Reassignments and departures in the safety team underscore potential gaps in institutional continuity for hazard assessment.
- OpenAI’s willingness to adjust safety rules based on competitors’ actions raises questions about industry-wide norms and incentives.
Key facts
- OpenAI is hiring a Head of Preparedness responsible for executing its Preparedness Framework.
- Sam Altman said models are presenting challenges including potential impacts on mental health and discoveries of critical security vulnerabilities.
- The role includes tasks such as enabling cybersecurity defenders without empowering attackers, managing biological capability releases, and building confidence in running self-improving systems.
- OpenAI announced a preparedness team in 2023 to study risks from immediate threats like phishing to more speculative ones like nuclear misuse.
- Aleksander Madry, the previous Head of Preparedness, was reassigned to focus on AI reasoning within the past year.
- Other safety executives at OpenAI have left the company or moved into roles outside preparedness and safety.
- OpenAI updated its Preparedness Framework recently and said it may adjust safety requirements if competitors deploy high-risk models without similar protections.
- Recent lawsuits allege ChatGPT amplified users’ delusions, worsened social isolation, and in some cases was linked to suicides; OpenAI says it is working to improve the model’s ability to spot emotional distress and connect users to support.
What to watch next
- Whether OpenAI fills the Head of Preparedness role and who is appointed — not confirmed in the source.
- How the new leader will restructure or reprioritize the preparedness team after recent reassignments and departures — not confirmed in the source.
- If and how OpenAI will change safety requirements in response to rival labs releasing models it deems "high-risk" — the framework says it might, but concrete actions and timing are not confirmed in the source.
- Outcomes of ongoing legal actions and whether they trigger product or policy changes at OpenAI — not confirmed in the source.
Quick glossary
- Preparedness Framework: A structured approach used by an organization to identify, monitor and plan responses to emerging capabilities that could pose severe harms.
- Generative AI: A class of artificial intelligence models that produce content such as text, images or code in response to prompts.
- Vulnerability: A weakness in software, systems or processes that could be exploited to cause harm or unauthorized access.
- Self-improving system: A system that can autonomously modify its own behavior or internal parameters to enhance performance, which may raise novel safety considerations.
Reader FAQ
Why is OpenAI hiring a Head of Preparedness?
To lead execution of its Preparedness Framework and study frontier AI risks spanning areas like cybersecurity, biological capabilities and mental health.
What happened to the prior Head of Preparedness?
Aleksander Madry was reassigned to an AI reasoning role within OpenAI; other safety executives have also shifted roles or left.
Will OpenAI change its safety requirements?
The company updated its Preparedness Framework and said it might adjust safety requirements if a competitor releases a high-risk model without similar protections, but specific changes are not detailed.
Are there legal claims related to mental-health harms from OpenAI’s products?
Recent lawsuits allege harms including reinforced delusions, increased isolation, and suicides; OpenAI says it is working to improve the model’s ability to detect emotional distress and connect users to support.
OpenAI is looking to hire a new executive responsible for studying emerging AI-related risks in areas ranging from computer security to mental health. In a post on X, CEO Sam…
Sources
- OpenAI is looking for a new Head of Preparedness
- Head of Preparedness
- OpenAI Is Hiring Head of Preparedness, Amid AI …
- Sam Altman is hiring someone to worry about the dangers …
Related posts
- Claude Code creator says Claude wrote all his code for the last month
- Sam Altman hires Head of Preparedness to tackle AI risks, including mental health
- Year in Review: How AI Is Reshaping Police Reports and Oversight