An algorithmic audit examines an AI system to check whether it works as intended, whether it produces biased or discriminatory outcomes, and whether it meets legal and ethical standards. Think of it as the AI equivalent of a financial audit - an independent review that goes beyond what the developers themselves have tested. Impact assessments take a broader view, examining the potential effects of an AI system on individuals, communities, and society before it's deployed. The EU AI Act mandates fundamental rights impact assessments for high-risk systems, and similar requirements are emerging elsewhere. In practice, auditing AI is harder than auditing financial statements. Models can behave differently across populations, their performance can drift over time, and the data they were trained on may no longer reflect reality. There's also a shortage of qualified auditors and a lack of consensus on methodology - what exactly should an audit measure, and against what standard? Despite these challenges, algorithmic audits are becoming a core part of responsible AI deployment. If you're building or buying AI systems, you should be planning for how they'll be audited, who will do it, and how often.