The body, established following the inaugural AI Safety Summit in 2023, was tasked with creating measures businesses can use to test AI systems before public launch, in line with President Biden’s Executive Order on AI safety.
After Trump revoked his predecessor’s policy back in January, the President’s demand for reform will see the AI Safety Institute become the Centre for AI Standards and Innovation (CAISI).
The body, which will reside with the Commerce Department, will work with federal agencies to create evaluations and assessments of US and “adversary” AI systems.
The Commerce Department contends that changes to the new look CAISI will not, however, stifle innovation.
“For far too long, censorship and regulations have been used under the guise of national security,” said Howard Lutnick, the US Secretary of Commerce. “Innovators will no longer be limited by these standards.”
“CAISI will evaluate and enhance US innovation of these rapidly developing commercial AI systems while ensuring they remain secure to our national security standards.”
The revamped body will work closely with the National Institute of Standards and Technology (NIST) to develop guidelines and best practices to measure and improve the security of AI systems.
CAISI will work to create voluntary agreements with private sector AI developers and create evaluations of AI capabilities that may pose risks to national security.
The body will also be tasked with evaluating issues with the potential adoption of “foreign AI systems” as well as the state of international AI competition, with President Trump recently claiming the US was “way ahead of everybody at AI”.
The reworking of the institute comes as President Trump attacks what his administration perceives as ‘ideological bias’ from foundation models, with both the White House and the Republican-led Congress repeatedly lambasting “woke AI” — echoing previous anger towards perceived algorithmic discrimination, such as search results from platforms like Google.
The President has already slashed the Biden-era AI Executive Order, describing it as “harmful” and that it “imposes radical left-wing ideas on the development” of AI technologies.
Earlier this year, Wired reported that staff at the body formerly known as the AI Safety Institute were instructed to prioritise reducing ideological bias in systems they were evaluating.
Staff were also encouraged to focus on ‘America First’ efforts, including testing tools to “expand America’s global AI position”.
