The EU is facing Big Tech pressure to revamp its AI law – POLITICO

Brilliant news from Brussels: the European Union is being lobbied to revise its AI rules by the very companies those rules exist to cover, which is roughly the arrangement you'd get if you let the fox chair the henhouse refurbishment committee. POLITICO reports that Big Tech pressure is bearing down on the law, with Microsoft's Brad Smith and Google DeepMind among the names attached to the effort. Meanwhile, EU member states — Hungary among them — have raised the charmingly foundational question of whether anyone actually knows when models are being trained and tested. It appears the answer may be no, and the authorities would quite like some paperwork.

The member states have also posed the delightful follow-up of whether a company can be held liable if its agents do harm — the sort of question you'd hope someone had settled before the agents were let out of the building. Czech government AI envoy Lukáš Kačena has called the EU's AI rules a 'great start', which is the phrase an envoy deploys when he wants to be quoted warmly while privately eyeing the exits. 'Great start' is also what you say about a house with no roof, a cake with no flour, or a lifeboat with most of the bottom missing. Mr Kačena says the rules will need updates; the member states are proposing action at Union level. Everyone agrees something must be done, and nobody has volunteered to define what.

So the sequence is now clear enough. The capability ships, the agents get deployed, and then the question of liability arrives as a follow-up item on a Tuesday. The pressure to revamp the law comes not from citizens who have watched an agent do harm but from the companies best positioned to profit from the revamp, which at least has the virtue of being legible. 'Great start', we are told. A start implies a middle and an end, and nobody in this story has any intention of arriving at either.