Keelforge

Screen 03 · R-007 · Layers 05 & 07

AI assurance against the EU AI Act

Section 4.3 Governance · High weight · Owner P. Adeyemi · BID-2026-014

Red

Requirement text

Describe your AI assurance approach against the EU AI Act, including GPAI classification, risk tiering, conformity assessment and post-market monitoring.

Drafted answer

draft.v3 · Drafting Agent

ClaimThe AI assurance approach classifies every model in scope against the EU AI Act GPAI thresholds, tiers each model against the Act's risk categories, applies conformity assessment appropriate to the tier and runs post-market monitoring with quarterly reporting to the buyer's model risk committee.

Evidence anchorEU AI Act GPAI assessment note EV-2026-0195 · Model inventory EV-2026-0196 · Post-market monitoring runbook EV-2026-0197.

ConfidenceMedium-high. GPAI classification and tiering are proven; conformity assessment at the highest tier requires a named conformity body which is buyer-selectable.

Counter-viewA competitor bidding under a lighter interpretation of the Act (for example classifying as limited-risk when the correct classification is high-risk) will show a leaner cost line; the response should lead with the correct classification and treat the cost delta as a governance premium.

Evidence chain

Layer 07
  • EU AI Act GPAI assessment note EV-2026-0195

    storage://evidence/EV-2026-0195.pdf

    signed_by P. Adeyemi · 13 Aug 2026 · 6e22b8…04af

  • Model inventory EV-2026-0196

    storage://evidence/EV-2026-0196.csv

    signed_by J. Okonjo · 12 Aug 2026 · cc7314…b920

  • Post-market monitoring runbook EV-2026-0197

    storage://evidence/EV-2026-0197.pdf

    signed_by H. Lindqvist · 10 Aug 2026 · 1902ef…7ac5

Discriminator injection

Layer 16

Against a competitor known to classify GPAI as limited-risk, sharpen to we have classified this as high-risk and the buyer's regulator will thank us for it.