US seeks ‘unquestioned’ AI dominance

Red Tape

The US will remove red tape, which the government believes is holding the private sector back.

Trump made some moves already in this regard, such as rescinding an executive order (EO).

Former president Joe Biden issued a directive in 2023 which attempted to set standards on AI safety and security and privacy, and encourage responsible development. The new administration believes that Biden’s EO was “a foreshadowing of an onerous regulatory system”, and many of this week’s actions reflect a general trend to undo his legacy.

The plan also outlines actions to encourage the adoption of open-source and open-weight AI models in critical sectors like healthcare, empower initiatives for expanding AI enabled education, and prepare the workforce to its impact. The plan’s second pillar proposes a variety of actions, including streamlining the building and planning processes for datacentres and chip fabrication facilities as well as expanding and preserving the vast energy infrastructure needed to run AI models.

The third pillar outlines plans to export AI technology to other countries.

They should not turn to rival countries like China. The plan includes initiatives to counter China’s influence in international governance bodies.

The AI action plan also contains provisions for more AI focused cyber initiatives. These include the potential creation of a AI-specific Information and Analysis Centre (ISAC) within the Department of Homeland Security.

There is no ‘woke’ AI at the White House

In addition to the AI action plan, and in line with its disavowal of diversity, equality and inclusion (DEI), the Trump administration launched a broadside on so-called ‘woke’ AI.

Text of

A second EO was signed by Trump on the 23rd of July and states that large language models should be neutral, nonpartisan tools. “They should not manipulate responses to favour ideological dogmas like DEI”.

According to the EO, in an AI context this could mean targeting topics like representation of ethnic or homosexual minorities, critical racial theory, discrimination based on race or sex or intersectionality, as well as transgender rights, unconscious bias, and unconscious bias. The EO’s guidelines apply to models used by the federal government but will likely have ramifications that go beyond it.

Broadly supporting

Matt Mittelsteadt Research Fellow at Washington DC based libertarian think-tank the

Cato Institutewas generally supportive of the plan’s goals.

On the positive sideand the order includes an emphasis on innovation that is much needed, particularly the forthcoming effort to analyze federal rules in order to identify onerous regulation,” he said. “Also noteworthy is the renewed focus on open source and related initiatives to enable open-source developers.

Mittelsteadt said that the stated emphasis on AI diplomacy may be crucial, as America’s AI success is dependent on foreign market access. “Finally the increased emphasis on cyber security, and specifically enabling a cyber defence, is matched with the current state of risks.”

He pointed out a few sticking points, including the use of government procurement to shape model output according to ideological lines.

“Not just is ‘objectivity,’ elusive from a philosophical perspective, but efforts to contain perceived bias have not worked,” said Mittelsteadt.

This is also a mistake in terms of competitiveness. If this policy is successful in shaping American models, then we will lose international clients who won’t want to buy models that are shaped by the whims of a foreign government.