A legal framework for AI and its use of data should maximize benefits to be drawn out of the use of AI and big data and minimize its damages for society and individuals.
The general principle that should guide data protection laws is that data ownership lies with the individual as default. The individual can allow the usage or block the usage of his/her data. Deleting data is possible at any time.
Focus areas for minimizing damages:
- Forbid trading of critical personal data (health, personal economic data, political, religious or sexual orientation, nutrition?) between organizations
- Regulate "data silos": personal data might be collected for certain uses but cannot be used for other sensible purposes within the same company. Data storage must be separated (e.g. use of data in insurances)
- AI Label: All processes where AI takes decisions should be transparently labelled.
- Regulate and control level of data storage safety: physical safety (where is storage?) and prevention of hacking/possible abuses.
- -> similar to nowadays motor vehicle inspection, data storage for AI systems needs to be checked and certified every 2 years.
Focus areas for maximizing benefits:
- Swiss government should initiate and legislate a neutral organization responsible for beneficial and safe use of collected data --> free access for research organizations to crucial data for development of beneficial AI.
- The organization could be a foundation or association, similar for example to the SRG media --> financed by government but not government or parliament controlled.
- Free access to data for development of beneficial AI is decided by an ethics commission
- the neutral organization could potentially licence the use of anonymized personal data for business purposes
- Collected crucial data could be:
- Patient and treatment data of public hospitals
- Health insurance data
- criminal data records
- data of the Swiss Federal Statistical Office
- Data storage should be in a physically safe location (fortification in Swiss alps?)