Principal Ethicist / AI Ethics Manager

Evaluating the utility and performance implications of different fairness and XAI libraries for enterprise-wide adoption. You won't be writing daily code, but you'll guide your team and make architectural decisions.

Leading the selection, implementation, and enterprise-wide integration of an ML observability platform, defining the core governance workflows and custom monitor configurations for fairness metrics.

Owning the GRC module for AI, using it to generate board-level risk reports, manage the entire AI model inventory, and ensure compliance with regulatory requirements. You'll define how the platform is used.

Using tools like Notion or Miro for strategic planning, roadmap development, and designing new AI ethics review processes. You'll also oversee how your team uses Confluence and Jira for their day-to-day work.

Influencing data governance policies at the platform level to embed fairness and privacy by design. You'll understand how data flows and is managed, guiding technical teams on best practices.

Creating and presenting high-level risk dashboards to the C-suite and Board, translating complex ethical metrics into clear business impact and strategic recommendations.

• Product VPs and Directors
• Engineering Leads and Architects
• Legal and Compliance Teams
• Data Science and Machine Learning Leads
• Internal Audit and Risk Management
• The AI Ethics Board (AIEB)

• Industry Regulators (e.g., ICO, CMA)
• External Auditors
• Academic Researchers in AI Ethics
• Industry Consortia and Standards Bodies

The rapid evolution of large language models (LLMs) and generative AI brings entirely new ethical risks, from hallucination and misinformation to deepfakes and intellectual property issues. You'll need to be at the forefront of understanding and mitigating these.

Model Alignment & Value Alignment · Synthetic Data Ethics · Frontier AI Safety · Intellectual Property & Attribution for Generative AI

• This week: Read key research papers on LLM safety and alignment from organisations like OpenAI, Anthropic, or Google DeepMind.
• This month: Experiment with prompt engineering techniques for red-teaming generative models, trying to elicit harmful or biased outputs.
• Next 3 months: Lead an internal workshop for your team on the ethical risks and mitigation strategies for a specific generative AI use case relevant to our business.
• Next 6 months: Develop a draft policy for the responsible use of generative AI within your product portfolio, addressing issues like data provenance and output validation.

Quick win: Start using advanced generative AI tools (e.g., Claude, GPT-4) for brainstorming ethical scenarios, drafting policy summaries, or generating diverse test cases for bias detection. Get hands-on.

Beyond fairness, the robustness and security of AI systems are becoming critical ethical concerns. Adversarial attacks, data poisoning, and model inversion can lead to unreliable, unsafe, or privacy-compromising outcomes. You'll need to oversee the strategies to counter these.

Adversarial Machine Learning · Data Poisoning Attacks · Model Inversion & Membership Inference Attacks · Explainability for Security Audits

• This quarter: Familiarise yourself with common adversarial attack types (e.g., FGSM, PGD) and defence mechanisms.
• Next 6 months: Work with our security and MLOps teams to integrate robustness testing into our standard model validation pipeline.
• Next 12 months: Lead a 'red team' exercise focused specifically on adversarial attacks against one of our high-risk AI models.
• Ongoing: Stay updated on the latest research in AI security and robustness, and how it impacts ethical considerations.

Quick win: Review existing security protocols for our data pipelines and model deployment, identifying potential weak points that could introduce ethical risks through malicious intent.