This historical CHSPS news article was published in 2023 and documents the standards and program-development record presented for that year. For present requirements and status, use the current Standards Registry, Program Status and other controlling public records.
What changed in 2023
Consequential AI moved from experimentation into an explicitly governed lifecycle. The standards addressed AI-supported item development, scoring, proctoring and candidate services, then added human review, subgroup evidence, challenge rights and model, prompt, vendor and configuration-change revalidation.
Introduced explicit governance for AI-supported item development, scoring support, proctoring signals and candidate-service functions.
Expanded human review, subgroup evidence, challenge rights and automated-flag disposition controls.
Added model, prompt, vendor and configuration-change controls with monitoring and revalidation triggers.
2023 in context
The standards architecture begins treating consequential AI as a governed lifecycle rather than a generic software feature.
This work becomes the technical foundation for AIG, HSAI-AI, APS, ALT and ALTP in the following years.
The important change is not simply that another document number appears in the register. Each controlled edition represents a defined point at which requirements, scope language, evidence expectations or review controls were made more explicit. Program additions follow the same logic: a new pathway is introduced when the subject requires a distinct unit of accreditation or certification, not merely because the portfolio can support another name.
Controlled editions issued during 2023
The edition sequence below shows how the standards architecture developed during the year. The sequence is cumulative: later editions within the same year supersede the earlier controlled version while preserving the historical change record.
AI-assisted assessment controls
Introduced explicit governance for AI-supported item development, scoring support, proctoring signals and candidate-service functions.
The revision is part of the YYYY.N controlled-edition sequence: the year identifies the publication cycle and the number after the period identifies the sequence of controlled editions within that year. A later sequence number represents a substantive controlled revision, not a decimal software version.
Human oversight, fairness and re-review
Expanded human review, subgroup evidence, challenge rights and automated-flag disposition controls.
The revision is part of the YYYY.N controlled-edition sequence: the year identifies the publication cycle and the number after the period identifies the sequence of controlled editions within that year. A later sequence number represents a substantive controlled revision, not a decimal software version.
AI change control and revalidation
Added model, prompt, vendor and configuration-change controls with monitoring and revalidation triggers.
The revision is part of the YYYY.N controlled-edition sequence: the year identifies the publication cycle and the number after the period identifies the sequence of controlled editions within that year. A later sequence number represents a substantive controlled revision, not a decimal software version.
Program-development position in 2023
The standards architecture begins treating consequential AI as a governed lifecycle rather than a generic software feature.
No additional current portfolio code is assigned to this year in the chronology. The work instead strengthened the common controls that later program standards rely on. That distinction matters because adding a program before its scope and evidence model are mature would make the portfolio less coherent, not more complete.
How this year connects to the current 16-code portfolio
The current CHSPS portfolio is intentionally layered. general accreditation pathways address certification programs, education providers, courses, assessment-based credentials and work-based learning. Specialty accreditations address assessment integrity, remote examinations, medication safety, workforce competency, clinical education and AI governance. Supplements, system certifications and scope modules then address narrower uses such as AI-assisted assessment, AI proctoring technology, AI learning platforms, provider AI-learning use and telehealth workforce competency.
This work becomes the technical foundation for AIG, HSAI-AI, APS, ALT and ALTP in the following years.
Evidence and decision principles carried forward
Across the chronology, four principles remain stable. First, scope has to be defined before review. Second, a requirement must be supported by controlled documentation and implementation evidence. Third, review and final decision authority must remain impartial and traceable. Fourth, a public claim must stay within the approved scope and remain independently verifiable. New program standards add subject-specific requirements, but they do not replace these core controls.
What readers should use today
Historical editions explain how the architecture developed. They do not replace the current controlled standards. Applicants and accredited organizations should use the current Standards Library, program page and written CHSPS scope confirmation for present requirements. Public status should be checked through the current verification record.
More 2023 coverage
Related CHSPS articles published for 2023, each tied to the year's controlled standards and program record:
- AI-assisted assessment — How generative and predictive tools enter item development, scoring support and candidate services.
- AI change control — Why model, prompt and vendor changes require revalidation and monitoring.
- Automated proctoring fairness — Why anomaly flags should be evidence rather than final misconduct decisions.
- Human oversight — What meaningful human review requires when automated recommendations affect people.