One of the most significant security updates to ChatGPT is now rolling out — and OpenAI is actively seeking a security engineering professional to take Lockdown Mode to the next level. As the company tightens access to external resources and restricts certain product capabilities, it is simultaneously growing its team to build and refine these defenses. This role aims to harden ChatGPT against prompt injection and data exfiltration risks, making the platform safer for enterprise and personal users dealing with highly sensitive information. According to a recent report covered by Help Net Security, the new Lockdown Mode is an opt-in setting available across Free, Plus, Pro, and self-serve ChatGPT Business plans. It targets organizations and individuals who require robust protection against adversaries attempting to extract proprietary or confidential data through carefully crafted prompts. OpenAI’s decision to staff this feature separately underscores its commitment to responsible AI — and the critical need for talent that can bridge product security, large language model (LLM) safety, and user experience.
About the Role
This position involves designing, implementing, and maintaining the core security mechanisms that power Lockdown Mode inside ChatGPT. The role requires close collaboration with product, research, and infrastructure teams to ensure that restrictions on external resources and internal tooling are both airtight and minimally disruptive to legitimate workflows. The successful candidate will help define how the model handles unverified requests, audits behavior for injection signals, and evolves the lockdown posture as new attack vectors emerge.
Key Responsibilities
- Architect and develop the Lockdown Mode control plane, including access policies for external APIs, browsing, and code interpretation features.
- Research and model novel prompt injection techniques and build detection and mitigation logic within the ChatGPT runtime.
- Work with LLM safety researchers to fine-tune guardrails that prevent data leakage without degrading model performance.
- Integrate Lockdown Mode with authentication, session management, and encryption layers for end-to-end protection.
- Conduct threat modeling and red-team exercises specific to data exfiltration via conversational AI interfaces.
- Collaborate with enterprise sales and customer success teams to tailor Lockdown Mode for regulated industries such as finance, healthcare, and legal.
- Document security best practices and contribute to public-facing technical guides that educate users on when and how to activate the feature.
Requirements
- 5+ years of experience in application security or security engineering, with a focus on SaaS or cloud-native platforms.
- Deep understanding of web security fundamentals: authentication protocols, API security, CSP, and sandboxing techniques.
- Proficiency in Python and familiarity with LLM inference pipelines; experience with LangChain or similar orchestration frameworks is a plus.
- Hands-on knowledge of common injection attacks (SQL, command, prompt) and their mitigations in AI/ML environments.
- Track record of shipping security features or tools in a product-oriented engineering team.
- Excellent communication skills, with the ability to explain complex security tradeoffs to both technical and non-technical stakeholders.
Compensation & Benefits
- Competitive base salary and equity packages commensurate with experience.
- Comprehensive health, dental, and vision coverage for employees and their families.
- Flexible work schedule and the choice to work remotely or from the San Francisco headquarters.
- Generous learning and development budget, including conference attendance and certification support.
- Parental leave, mental health support, and wellness stipends.
- Access to state-of-the-art compute infrastructure and collaboration with world-class AI researchers.
How to Apply
Candidates interested in shaping the future of secure AI can use the Apply Now button on this page. The link will redirect to the official OpenAI careers portal, where full role details and application instructions are available.