Glossary · 6 · Quality, risk and governance
Accountability (AI)
Also known as: Accountability, AI accountability, Answerability
German: Rechenschaftspflicht
In AI governance, accountability is the assignment of a named person, role or organization that answers for a decision, output or system — including the duty to explain it, correct it and bear its consequences. It presupposes intent, authority and the ability to act, so it rests with people and organizations rather than with a model or an agent, however autonomous the system appears. <cite index="7-1,7-4">The OECD AI Principles put it as a duty of organisations and individuals that develop, deploy or operate AI systems to be held accountable for their proper functioning, based on their roles, the context and their ability to act.</cite> In practice, accountability becomes visible only when it is documented: a role, a review step and a record that shows who decided what.
- Intermediate
- Technical writers
- Technical marketers
- Technical project managers
- Developers
In one sentence
Accountability in AI explained: who answers for an AI output, how to assign it, and why it stays with people.
Example
Before an AI-drafted troubleshooting chapter ships, the lead technical writer signs it off in the review tool, so an audit can show a named person who answers for the instructions.
How it applies
- Name a person, not a team: Frameworks ask for structures, not good intentions — the NIST AI Risk Management Framework states under GOVERN 2 that accountability structures are in place, and its playbook expects roles, responsibilities and lines of communication for mapping, measuring and managing AI risks to be documented and clear to individuals and teams. In a documentation project that means one owner per deliverable, not "the AI wrote it".
- Write accountability into the workflow: Pair each AI-assisted step with a review gate and a record of who approved it. A human in the loop without a signature leaves no evidence; an AI usage policy should say which tasks require review and who signs off.
- Technical writers: You answer for published instructions even when a model drafted them — check facts against the source of truth, since hallucination is a property of the tool, not an excuse. Safety-relevant content needs expert sign-off on the record.
- Technical marketers: Claims, images and localized copy stay your responsibility, including rights clearance and AI content disclosure where your policy or market requires it.
- Technical project managers: Map accountability per lifecycle phase — vendor, integrator, deployer, reviewer — and keep it in the RACI, the contract and the risk register. Under the EU AI Act, Article 26 requires deployers of high-risk AI systems to take appropriate technical and organisational measures to ensure they use such systems in accordance with the instructions for use, and to assign human oversight to natural persons who have the necessary competence, to keep logs and to inform providers and authorities of risks. Naming the Act, a standard or this entry never establishes compliance by itself; only your implemented and documented measures can do that.
- Developers: Build the evidence trail — log prompts, model and version, tool calls and human approvals, so an output can be traced back to a decision. Accountability for code produced with an AI coding assistant or an AI agent stays with the committing engineer and the reviewer.
- Make it auditable: Evaluation (evals) results, a model card and change records turn a claim of diligence into demonstrable practice — the same logic as the GDPR's accountability principle in Article 5(2), read together with Articles 24 and 32, under which the controller is responsible for and must be able to demonstrate compliance with the data protection principles.
Accountability vs. responsibility
Responsibility is about doing the work: a writer drafts, a developer integrates, a reviewer checks. Accountability is about answering for the result — justifying it, fixing it and carrying the consequences — and it cannot be split across a crowd without disappearing. Tasks can be delegated to a tool, a supplier or an agentic workflow; the duty to answer cannot. This is why Responsible AI programs fail when they list values but never name who signs.