Shareholder agreements can turn regulatory obligations into contractual commitments.
The EU AI Act imposes governance, transparency, and risk management duties on AI providers.
These duties include providing technical documentation and assessing systemic risks.
Such regulatory clarity can attract international investors to European AI companies.
Examples
- Technical documentation for General Purpose AI (GPAI) providers.
- Transparency regarding training data content for GPAI providers.
- Systemic risk evaluation for high-impact AI models.
Assuming shareholder agreement clauses can genuinely enforce AI Act compliance for general-purpose AI seems like a major oversight, because what happens when the actual, technical risks are missed?
It's a huge downside if you think a piece of paper will stop an AI model from going sideways; we'd need detailed audits and constant oversight, not just a contract.
For example, if an AI starts generating biased outputs, a shareholder agreement won't fix the underlying algorithmic flaw or prevent the reputational damage.
We need proper fail-safes and regulatory thresholds that are technically robust, not just legal declarations.