Maintain technical documentation and publish: Under the EU AI Act, which obligation applies to it as a
Every GPAI provider must maintain technical documentation and publish a sufficiently detailed summary of the training content, independent of any systemic-risk designation.
The question
A company places a general-purpose AI (GPAI) model on the EU market but the model is not designated as posing systemic risk. Under the EU AI Act, which obligation applies to it as a provider of a GPAI model?
Preparing for AIGP? Take the free 5-min readiness quiz →
- Perform state-of-the-art adversarial evaluation and promptly report every serious incident directly to the AI Office.Plausible but wrong: adversarial model evaluation and serious-incident reporting are additional obligations reserved for GPAI models with systemic risk, not the baseline for every GPAI provider.
- Register in the EU high-risk database and complete a full third-party conformity assessment before release.Plausible but wrong: database registration and conformity assessment are high-risk AI system obligations, a separate regime from the GPAI-model rules in Chapter V.
- Maintain technical documentation and publish a sufficiently detailed summary of the content used to train the model. ✓Correct: all GPAI providers must keep technical documentation, pass information to downstream providers, adopt a copyright policy, and publish a sufficiently detailed training-data summary, regardless of systemic risk.
- Secure explicit opt-in consent from every individual whose personal data appears in the training set.Plausible but wrong: this conflates the AI Act with a GDPR-style consent requirement; the Act's GPAI baseline requires a training-data summary and copyright policy, not per-individual consent.
The trap
Believing systemic-risk or high-risk obligations apply to every GPAI model rather than only the baseline documentation and training-summary duties. How to remember it
Every GPAI provider must maintain technical documentation and publish a sufficiently detailed summary of the training content, independent of any systemic-risk designation.
How many of these would you get right?
One of 1581 AIGP questions on Certsqill. Take a free five-minute check and see your score per domain — not one number, but which section to open tonight.
Test your AIGP readiness — freeMore Understanding How Laws, Standards and Frameworks Apply to AI questions
- Transparency and provision of information: Which requirement does this most directly implement? →
- Placing on the market an AI system that breaches: Which category of violation is subject to the highest tier →
- Deployer - it must operate the system per the instructions: Which role does the hospital hold under the EU AI →
- All 394 Understanding How Laws, Standards and Frameworks Apply to AI questions →
Part of the Certsqill AIGP question bank · Understanding How Laws, Standards and Frameworks Apply to AI ·
Every answer, right and wrong, comes with its own explanation.
