Developers of the “most powerful AI systems” in the US must share security tests and other “sensitive information” with the government. This is clear from the presidential order issued today by US President Biden.
Biden’s order is part of a series of moves by the US government to try to gain more control over AI (artificial intelligence). It specifically concerns companies developing AI systems that pose a “serious risk” to US security. It does not describe what kind of risk it involves. Additionally, it is not specified who decides when such a risk exists.
A presidential decree is not a law. The President needs Congress for that. So Biden is calling on lawmakers to come up with legislation.
The next generation of AI
insists an anonymous government official For the Financial Times The mandate mainly applies to next-generation AI systems, not for example GPT4, the engine behind the latest version of ChatGPT, an advanced text generator.
Although developments are moving quickly, the question is when companies will have to deal with these rules in practice, and it suggests that the US does not want to be too harsh on its own companies with these new rules.
The White House and tech giants have been debating AI for some time. In July, the most prominent AI players committed to non-binding agreements. This includes a promise to share information about the security and reliability of their systems. With Biden’s decision, that lack of commitment disappears.
Watermark for AI-generated work
In addition to the requirements for large tech companies, it’s about creating a watermark so that work becomes clear as AI is developed. There’s a limitation here: it can only be used by central governments, but companies don’t have to adopt it. The White House hopes to set an example for companies and governments around the world.
This includes protecting the privacy of Americans, promoting equality, and protecting the rights of consumers, patients, and students and employees. The order also ensures that the US government uses AI in a responsible and effective manner.
The Biden administration’s moves come on the same day as a voluntary code of conduct for AI launched by the G7 nations and the European Union, Reuters reported yesterday. It sets out how key countries want to deal with AI against the backdrop of privacy and security concerns.
AI Summit United Kingdom
A two-day AI meeting will begin on Wednesday in the United Kingdom, where its focus will be on security. Vice President Harris has arrived there on behalf of the United States. Commission President Van der Leyen is expected to represent the EU and State Secretary Van Hafelen will represent the Netherlands. A delegation from China is also expected.
“Passionate analyst. Thinker. Devoted twitter evangelist. Wannabe music specialist.”