Software Engineer, Safeguards (London)

Anthropic is seeking a Software Engineer for their Safeguards team in London, focusing on building critical safety and oversight mechanisms for AI systems. This role sits at the intersection of AI safety and security engineering, working to ensure AI systems remain reliable and beneficial.

The position involves developing sophisticated monitoring and abuse detection systems to maintain the integrity of Anthropic's AI platforms. Key responsibilities include building real-time defense mechanisms, analyzing user behavior patterns, and implementing automated enforcement actions when necessary. The role requires collaboration with research teams to improve model safety at the training stage.

The ideal candidate will bring 3-8+ years of software engineering experience, with particular emphasis on integrity, spam, fraud, or abuse detection systems. Strong technical skills in Python, SQL, and data analysis are essential, as is the ability to communicate complex technical concepts to diverse stakeholders.

Anthropic offers a competitive compensation package ranging from £240,000 to £325,000 GBP annually, along with benefits including equity options, parental leave, and visa sponsorship for international candidates. The company maintains a hybrid work model requiring at least 25% in-office presence.

As a public benefit corporation, Anthropic is committed to developing safe and beneficial AI systems. The company values diverse perspectives and encourages applications from candidates of all backgrounds, even if they don't meet every listed qualification. The role offers an opportunity to work on meaningful challenges in AI safety while contributing to the development of trustworthy AI systems.

The position combines technical expertise with the mission-critical goal of ensuring AI system safety and reliability. Successful candidates will join a collaborative team working on large-scale research efforts, contributing to the advancement of steerable, trustworthy AI technology. The role offers significant impact potential in shaping the future of AI safety mechanisms and oversight.