US and 17 other countries agree on recommendations for safe AI systems

Dado Ruvić/Reuters

The US, the UK and more than a dozen other countries have unveiled the first international agreement with recommendations for creating safe artificial intelligence systems. AI developers are encouraged to design their solutions to protect users and society from abuse.

Here's What We Know

The 20-page document calls for companies to implement measures such as monitoring AI systems for breaches, protecting data from tampering, and vetting software vendors.

According to Jen Easterly, director of the US Cybersecurity Agency, the number of countries supporting the idea of prioritising security in AI is important. This is the first step toward a harmonised position on the need for a responsible approach to the creation of new technologies, she said.

The agreement is another global initiative to regulate AI development. The UK previously held a summit on the topic and the EU is discussing legislation in this area.

The document does not address ethical issues of technology use or the problems of data collection for training. Nevertheless, experts consider it an important first step towards a unified approach to regulating AI globally.

Source: The Guardian