The US is on the verge of weakening a global AI treaty by exempting its largest tech companies from the requirements. Diplomats are meeting in Strasbourg, France, this week to finalise a comprehensive international agreement to better safeguard human rights in AI systems. The treaty has been negotiated for almost three years and responds to growing worries about the threats to fundamental human rights posed by AI technologies used in popular applications such as OpenAI's ChatGPT.

Why does it matter?

The text, drafted by the Council of Europe, a 46-country international organisation that oversees a major human rights convention, would create fundamental duties to safeguard human dignity, the rule of law, and democratic norms when AI is deployed. A final draft could be agreed upon by Thursday, with countries ratifying it subsequently.

However, the deal faces diplomatic opposition from the US, supported by Canada, Japan, and the UK, seeking to exclude private companies from its reach. The US, Canada, and Japan are non-voting observers at the Council of Europe, but without their signatures on the treaty, its power and reach would be watered down.

Washington is pushing for its firms to be exempt from the new rules on AI and human rights, sparking tensions in the negotiation process. The concerns have been more widely shared by civil society and academia. In January, an open letter sponsored by the non-profit Centre for AI and Digital Policy (CAIDP) and signed by hundreds of AI experts urged negotiators not "to abdicate our rights."

All this is happening as members of the European Parliament passed the landmark Artificial Intelligence (AI) Act on 13 March, making the EU the first major political organisation to set comprehensive rules for emerging technology.