It was announced yesterday that the outgoing secretary of state for digitisation is working on a draft proposal to ban civil servants from using generative AI software. It seems logical; a ban for civil servants. After all, no one wants reports containing sensitive data to leave an organisation asking "please summarise this".
Generative AI is a form of artificial intelligence that discovers patterns in large amounts of training data and can generate something new based on them with similar characteristics to the training data. Some of the training data consists of copyrighted books and newspaper articles, and there is no explicit judgment yet from case law on whether that is allowed. This is one of the concerns raised in the draft proposal by the secretary of state.
A legally conclusive answer to the question of to what extent an AI application violates copyright, and whether algorithms can qualify as 'authors', is likely to be years away. A ban on the use of AI software by civil servants until an irrevocable judgement is made seems unrealistic and certainly not future-proof. Indeed, the real objection is that legislation is falling further and further behind technical developments.
Also, a generative AI application can derive highly sensitive information from user interaction, writes the Volkskrant, which is undoubtedly true. However, this privacy argument AI software does not only apply to AI applications. Questions asked to search engines are also identifying (just search for "dog that urinates on everything"). Besides, everything in this objection points to an uncomfortable situation.
Virtually every ministry, political party, newspaper or government agency, sends their visitors to Facebook, X, Instagram or Whatsapp. These are all commercial companies that can "derive highly sensitive information from user interaction".
The last argument mentioned in the Volkskrant is that programmes like ChatGPT "can be used to make a decision about a person". That objection seems the least valid. Dice can also be used to make a decision about a person. The point is that technological developments are happening at lightning speed.
We need to make people (and not just civil servants) aware of the advantages and disadvantages it has but also of the responsibility it entails. That requires a clear government vision of AI, and a constantly evolving training programme. Fortunately, that is also part of the draft proposal.
