The American President Joe Biden announced an agreement between major companies engaged in the development of artificial intelligence. The companies are aAmazon, Anthropic, Google, Inflection, Meta, Microsoft and OpenAI.
The agreement was announced in a short speech Roosevelt room in which he also affirmed that this was only a first step and that it was his own Administration is attempting to finalize an implementing regulation on AI. In addition, the President said he will ask that congress To legislate to ensure that American technology leadership occurs in a context where “Responsible innovation”.
Under the agreement, the companies will subject new AI systems internal and external tests before her release. They will commission independent bodies to examine their systems for any problems security breachdiscriminatory tendencies or risks to American rights, health information, or safety.
company incl Anthropic and Flexion AIhave committed to sharing the information to improve risk reduction with governments, civil society and academia and to report on it vulnerability as soon as they appear. Leading AI companies will integrate virtual watermarks in the material they generate and provides a way to distinguish real images and videos from those created by computers.
Nick CleggMeta’s President for Global Affairs said these commitments are an “important first step in ensuring responsible boundaries are set for AI and a model for other governments to follow.”
“AI should benefit society as a whole. For this to happen, these are new powerful technologies must be built and used responsibly,” he said.
White House officials say those pledges are helpful balance The promises of artificial technology come with risks and are the result of months of intense activity lobbying backstage. Many of the leaders present at the White House attended a meeting with Biden and the vice president in May Kamala Harrisin which the administration warned the industry against liability guarantee the security of your technology.
“We need to make sure companies have their products tested pressure test during development and especially before release to ensure they don’t have unintended consequences such as B. a susceptibility to cyber attacks or use it to discriminate against certain people,” the White House chief of staff said in an interview Jeff Zients.
The guidelines do not require approval of specific external experts for the release of technologies and companies they are kept only to signal and not eliminate risks such as possible misuse or bias.
And other than public opinion, there are few mechanisms to enforce the use of technology social priorities like medicine and climate change.
“It’s a moving target,” Zients said. “So not only do we need to meet and implement those commitments, but we also need to understand what the next set of commitments are going to be.” technologies they will change”.
In discussions with experts, Biden has been warned that I algorithmic social media – like Facebook and Instagram by Meta and TikTok by ByteDance Ltd. – have already highlighted some of the risks that AI could bring. According to some sources, a city councilor advised the president to give the matter due attention and compare it with the Clone in the 1990s, so with the need for clear principles.
There White House She said she consulted with the governments of 20 countries before the announcement.
“I think all parties were ready or willing to move forward on this issue as quickly as possible because that’s how AI works: you can’t.” sleep on this technology,” said Deputy Chief of Staff Bruce Reed.
Also in Europe, where the EU AI law lies by far Superior In comparison to the law passed by the US Congress, the leaders recognized the need for voluntary compliance by companies before passing a binding law. A White House official estimated they might at least pass 2 years before European regulations start impacting AI companies.
At meetings with technology executives over the past three months, Thierry Breton, the European Union’s Internal Market Commissioner, has urged AI developers to join one “Compact for AI” set up some non-binding assignments.