Trustworthy AI EO (E.O. 13960)

Summary

Creates rules for government agencies to make sure they use artificial intelligence in a way that is safe, fair, and can be trusted by the public.

What problem does this solve?

The government is using more artificial intelligence, but there are no common rules to make sure it's used safely and fairly, which could make people not trust it. This order creates a set of principles for all agencies to follow, ensuring AI is used responsibly and builds public confidence.

What does this order do?

Establishes nine principles for trustworthy AI

Requires agencies to ensure their AI is lawful, purposeful, accurate, safe, understandable, responsible, regularly monitored, transparent, and accountable.

Requires agencies to create an inventory of AI uses

Mandates that each agency prepare and annually update a list of its non-secret uses of AI. This inventory must be shared with other agencies and made available to the public.

Excludes defense and national security AI

Specifies that these principles do not apply to AI used in defense or national security systems, which must follow their own separate guidelines.

Directs the Office of Management and Budget (OMB) to create AI guidance

Tasks the OMB with developing a public roadmap for creating or updating policies to support the government's use of AI, consistent with the order's principles.

Creates a plan to fix or retire non-compliant AI

Requires agencies to review their existing AI systems and develop plans to either make them consistent with the new principles or stop using them.

Establishes an AI track to recruit experts

Directs the Presidential Innovation Fellows program to create a special track to attract AI experts from the private sector and academia to work for the government.

Who does this affect?

Federal government agencies
The American public
Technology companies working with the government

What is the real world impact?

•

Builds public trust in government technology

Establishes clear, public principles for how the government uses AI. This helps ensure that AI is used in a way that is fair, safe, and respects American values, which can increase public confidence in government services.

•

Standardizes AI use across federal agencies

Creates a common set of rules for all non-defense agencies. This helps the government be more efficient and consistent in how it develops and buys AI tools, preventing each agency from creating its own separate standards.

•

Excludes high-stakes AI systems

Does not apply its principles to AI used for national security or defense. Critics might argue that the most powerful and potentially dangerous AI systems are exempt from this public-facing ethical framework.

When does this start?

This order became effective on December 3, 2020, and sets several deadlines for federal agencies to meet.

OMB policy roadmap

Within 180 days of the order (by June 1, 2021), the Office of Management and Budget must publicly post a roadmap for creating AI policy guidance.

Agency AI inventories

Within 180 days of the CIO Council's guidance, agencies must prepare their first inventory of AI use cases.

Public release of inventories

Within 120 days of completing their inventories, agencies must make them available to the public.

Plans for non-compliant AI

Within 120 days of completing their inventories, agencies must develop plans to fix or retire AI that does not meet the order's principles.

Recruitment of AI experts

Within 90 days of the order (by March 3, 2021), the Presidential Innovation Fellows program must establish an AI track to attract experts.