Securing AI systems and governing their use: threats, defences, frameworks, law. 13 pages. Generated by scripts/build_manifest.py - do not edit by hand.
Threat-model an AI feature before you build it by mapping what enters the model's context, what it can cause to happen, and whose authority each action runs under.
Know where every model weight, adapter, dataset and tool server came from, pin it, scan it and record it, the way you already do for code dependencies.
Understand why text in a model's context can act as an instruction, and why the fix is limiting what a persuaded model can do rather than hoping to block every attack.
Work out your role and risk tier under the EU AI Act and build a compliance program for the obligations that apply to you, on the timeline as amended in 2026.
Describe the system and who it affects, list what could go wrong across six risk categories, score likelihood and impact, decide controls and residual risk, and get a named owner to sign it off; this page gives you a copyable template and a worked example.
Responsible AI means turning six principles (fairness, reliability and safety, privacy and security, transparency, human oversight, and accountability) into concrete practices you can check, at every stage from deciding whether to use AI at all to monitoring it in production.
A working AI governance program is five things (named owners, an inventory of AI use, risk-tiered review, a few clear policies, and ongoing monitoring) that you can stand up in 90 days and then map onto NIST AI RMF, ISO/IEC 42001 or the EU AI Act as needed.