Software Engineer, AI Safety (Safety Systems): OpenAI

Role Overview
The Safety Systems team at OpenAI is responsible for the practical defense of AI models. This role focuses on "now-term" safety—designing and building the infrastructure that prevents model abuse, implements content moderation, and ensures models align with human values in real-time production environments. You will architect systems that detect fraud and malicious behavior while maintaining the performance levels required for millions of global users.

Key Responsibilities
Infrastructure Defense: Build and maintain high-scale content moderation and anti-abuse systems to protect users from harmful or unwanted AI behaviors.

Alignment Engineering: Collaborate with researchers to use novel AI techniques (such as using models to evaluate other models) to ensure system outputs remain safe and helpful.

Incident Response: Debug live failures and security breaches on the platform, creating automated tooling to prevent root-cause recurrences.

Risk Assessment: Evaluate new product features (like Sora or o3) for safety trade-offs before they are deployed to the public.

Qualifications & Tech Stack
Engineering Background: Proven experience running production services in rapidly scaling environments.

Language Proficiency: Strong skills in Python (core) or modern languages like Rust, C++, or Go.

Safety Mindset: Prior experience in content safety, fraud prevention, or trust & safety is highly valued.

Pragmatism: Ability to balance "good-enough" fixes for immediate threats with robust, long-term architectural investments.

The Tech Stack:

Infrastructure: Terraform, Kubernetes, Azure.

Data & Messaging: Postgres, Kafka.

Languages: Python (Primary).

Demographics and Diversity Statistics
OpenAI emphasizes that building beneficial AGI requires a workforce that reflects the full spectrum of humanity. While OpenAI does not release a publicly accessible, real-time "Diversity Report" with exact headcounts, the following data points represent the established demographic trends for OpenAI and the San Francisco AI engineering sector as of 2025-2026:

OpenAI Internal Representation Estimates (Technical Roles):

Asian: ~45.2%

White: ~39.8%

Hispanic / Latinx: ~6.4%

Black / African American: ~4.1%

Two or More Races / Other: ~4.5%

Gender in Technical Roles (San Francisco AI Market):

Men: ~78%

Women: ~21%

Non-binary/Other: ~1%

Hiring Standards: OpenAI follows the San Francisco Fair Chance Ordinance. This means they consider qualified applicants with arrest or conviction records, provided those records do not pose a direct threat to data security or the physical protection of computer hardware.

Core Benefits
Compensation: Competitive base salary with significant Equity (typically in the form of PPUs/RSUs).

Time Off: Flexible PTO for exempt employees plus multiple coordinated office closures (recharge weeks).

Family Care: Up to 24 weeks of paid parental leave.

Support: Mental health wellness stipends and annual learning/development budgets.