President Joe Biden will publish new national security memorandum Today on artificial intelligence, aimed at helping the American government to deploy AI and keep its advantage over China.
The United States must “ensure that our national security agencies adopt this technology in a way that aligns with our values,” said a senior White House official in the future before disclosure, adding that ‘A failure to do so “could put us in danger of a strategic surprise of our rivals, like China. “”
The memorandum provides advice – via a framework – on how to use AI for national security missions. “These requirements oblige agencies to monitor, assess and alleviate the risks of AI linked to invasions of privacy, prejudices and discrimination, the security of individuals and groups and other violations of the rights of the man, “according to an information sheet from the White House.
The directives will allow the government to better benefit from the new IA tools from Silicon Valley, said the White House.
“The innovation that occurred, in particular in this current wave of border artificial intelligencehas really been motivated by the private sector, and it is essential that we continue both to promote this leadership, “said the official. “AI is rooted in the premise that the capacities generated by the transformer And Great language model The revolution in AI, often called frontier IA, is ready to shape geopolitical, military and intelligence competition. »»
There are still major concerns concerning the application of these AI tools, such as the Openai Chatppt, to areas with high issues such as national security, given the propensity of these models to hallucinate and produce false positives . And data sets accessible to the public on which these models are largely formed may contain personal personal information (but legally obtained) on American civilians, as Defense one reported.
The national security community is “well aware” of the concerns, said the official. “We have to go through a system accreditation process. And it is not only for AI systems, but, you know, national security systems in general, “he said, pointing to the previous one Executive Decree on AI And the establishment of the IA security institute as two stages that the White House has taken to mitigate the risks of AI, in particular in the use of government.
The new memo designates the trade department IA security institute Like “the main port of contact in American industry in the American government”, according to the information sheet. These efforts are already joining the existing Ministry of Defense and Intelligence community Guidelines on the development and deployment of AI.
The manager has agreed that the availability of data from American citizens is an increasing problem. “We were very concerned about the ways in which the sensitive data of the Americans can be sold, really through the front door, collected first in bulk, then sold through data brokers, then we find ourselves between the hands of our opponents. And this is something that the president issued a decree on (in February) to try to restrict the opponent to some of this data. »»
But although the risks are real, the government must always establish a means for the national security community to “experiment” with AI through “pilots”, said the official.
“There will be challenges associated with the adoption of any new technology,” they said. “The frame … is the one that will be permanently updated.”
And these updates can be subject to political disturbances: the republican presidential candidate Donald Trump A juror to repeal Biden executive order on AI security.