Governance and accountability

Governance evidence

What agencies reported about the OMB M-25-21 minimum practices for high-impact AI. This is a disclosure analysis, not an audit, a compliance finding or an official assessment. An unanswered question is missing evidence, not evidence that a safeguard is absent.

Governance report

Practice-by-practice evidence

Denominator: 9 records · use cases that answered at least one practice question

Reported in placeReported in progressReported not in placeNot applicableOther or ambiguous answerNo answer reported

Pre-deployment testing

Pre-deployment testing
Reported in place: 4 (44.4%)Pre-deployment testing
Reported in progress: 3 (33.3%)Pre-deployment testing
Reported not in place: 1 (11.1%)Pre-deployment testing
No answer reported: 1 (11.1%)

AI impact assessment

AI impact assessment
Reported in place: 3 (33.3%)AI impact assessment
Reported in progress: 4 (44.4%)AI impact assessment
No answer reported: 2 (22.2%)

Independent review

Independent review
Reported in place: 4 (44.4%)Independent review
Reported in progress: 2 (22.2%)Independent review
Reported not in place: 1 (11.1%)Independent review
No answer reported: 2 (22.2%)

Ongoing monitoring

Ongoing monitoring
Reported in place: 3 (33.3%)Ongoing monitoring
Reported in progress: 3 (33.3%)Ongoing monitoring
No answer reported: 3 (33.3%)

Operator training

Operator training
Reported in place: 4 (44.4%)Operator training
Reported in progress: 2 (22.2%)Operator training
No answer reported: 3 (33.3%)

Fail-safe

Fail-safe
Reported in place: 3 (33.3%)Fail-safe
Not applicable: 4 (44.4%)Fail-safe
No answer reported: 2 (22.2%)

Appeal process

Appeal process
Reported in place: 1 (11.1%)Appeal process
Not applicable: 6 (66.7%)Appeal process
No answer reported: 2 (22.2%)

User and public consultation

User and public consultation
Reported in place: 6 (66.7%)User and public consultation
Reported in progress: 2 (22.2%)User and public consultation
Other or ambiguous answer: 1 (11.1%)

Practice	In scope	In place	In progress	Not in place	N/A	Other	No answer	Answered	In place of applicable
Pre-deployment testing	9	4	3	1	0	0	1	89%	44% of 9
AI impact assessment	9	3	4	0	0	0	2	78%	33% of 9
Independent review	9	4	2	1	0	0	2	78%	44% of 9
Ongoing monitoring	9	3	3	0	0	0	3	67%	33% of 9
Operator training	9	4	2	0	0	0	3	67%	44% of 9
Fail-safe	9	3	0	0	4	0	2	78%	60% of 5
Appeal process	9	1	0	0	6	0	2	78%	33% of 3
User and public consultation	9	6	2	0	0	1	0	100%	67% of 9

Answered = records with any answer ÷ records in scope. In place of applicable = reported in place ÷ (in scope minus not applicable, precluded and waived). Unanswered records stay in that denominator, so the rate is a lower bound on reported evidence, not an estimate of real practice.

Agency-by-safeguard heatmap

Cell: share reported in place, among applicable recordsIn place Answered

Agency by governance practice
	Testing	Impact assess.	Indep. review	Monitoring	Training	Fail-safe	Appeal	Consultation	n
SBA	33%	33%	33%	33%	33%	100%	100%	33%	3
EAC	0%	–	0%	–	–	–	–	100%	2
SSA	100%	100%	100%	50%	100%	100%	0%	100%	2
DOE	50%	0%	50%	50%	50%	0%	0%	50%	2

Hatched cells: the agency answered nothing for that practice, so there is no evidence either way. Compare agencies with care: each agency is one reporting unit, and portfolios differ in size and type.

Records requiring further investigation

7 records with at least one practice unanswered or in progressOpen in explorer

Use case	Stage	Test	AIA	Review	Monitor	Train	Fail-safe	Appeal	Consult
ChatGPT Enterprise DOE · DOE-498	Deployed	Reported in place	Reported in progress	Reported in place	Reported in place	Reported in place	Not applicable	Not applicable	Reported in place
OpenAI ChatGPT Pilot DOE · DOE-526	Pilot	Reported in progress	Reported in progress	No answer reported	No answer reported	No answer reported	Not applicable	Not applicable	Other or ambiguous answer
Internal Generative AI Tools for Staff Productivity EAC · EAC-2	Pilot	Reported not in place	No answer reported	Reported not in place	No answer reported	No answer reported	No answer reported	No answer reported	Reported in place
Routing emails from shared inboxes EAC · EAC-4	Pilot	No answer reported	No answer reported	No answer reported	No answer reported	No answer reported	No answer reported	No answer reported	Reported in place
Generative AI for SBA.gov product training SBA · SBA-13	Pre-deployment	Reported in progress	Reported in progress	Reported in progress	Reported in progress	Reported in progress	Not applicable	Not applicable	Reported in progress
ChatGPT for OMS Marketing SBA · SBA-15	Retired	Reported in progress	Reported in progress	Reported in progress	Reported in progress	Reported in progress	Not applicable	Not applicable	Reported in progress
Policy Assistance Tool (PAT) SSA · R3116	Pilot	Reported in place	Reported in place	Reported in place	Reported in progress	Reported in place	Reported in place	Not applicable	Reported in place

Unanswered analytical questions

Generated from the data; for follow-up with agencies

2 high-impact use cases mark an appeal process as not applicable. On what basis was each individual outcome judged not contestable?
5 governance answers are ambiguous (for example, an answer option from a different question). What did the agency intend?

These are questions the inventory raises but cannot answer. They are not findings of non-compliance.