The Language Access for All Act would allow federal agencies to use AI-assisted language tools but require qualified human verification, public transparency and recurring oversight.
U.S. Senators Andy Kim of New Jersey, Mazie Hirono of Hawaii and Kirsten Gillibrand of New York introduced the Language Access for All Act of 2026, or S. 4985, on July 15, 2026.
The legislation would prevent federal agencies from fully replacing qualified translators and interpreters with artificial intelligence or machine translation. Agencies could continue using automated language tools, but a qualified human would have to verify the resulting translation or interpretation.
The practical issue for federal buyers is therefore not simply whether AI can be used. It is whether that use is disclosed, authorized, verified by a qualified professional and supported by records that can be reviewed later.
The bill proposes a detailed accountability framework for AI-assisted federal language services, including annual technology disclosures, bias testing, privacy protections and recurring audits.
According to the official legislative record, S. 4985 was referred to the Senate Committee on Homeland Security and Governmental Affairs. It is the Senate companion to H.R. 7223, introduced in the House in January 2026.
Both measures remain proposed legislation. They have not passed Congress, and their compliance deadlines are not currently in effect.
The bill has already begun attracting attention within the language-services industry. For federal buyers, however, the more consequential question is what evidence agencies may need if it becomes law: documented human verification, disclosed system limitations and error rates, qualification records for assigned linguists and a reproducible audit trail.
Why Congress Is Reconsidering Federal Language Access
The legislation follows a major change in federal language policy.
Executive Order 13166, issued in 2000, directed federal agencies to improve access to their programs for people with limited English proficiency. It also instructed agencies providing federal financial assistance to develop Title VI guidance for their recipients.
President Donald Trump revoked that policy in March 2025 through Executive Order 14224, which designated English as the official language of the United States.
The order did not require agencies to discontinue all multilingual services. However, subsequent Department of Justice guidance directed federal agencies to reduce nonessential multilingual offerings and prioritize English-language education and assimilation.
The Language Access for All Act would place federal language-access requirements in statute, making them less vulnerable to changes between presidential administrations.
According to a KFF analysis of 2021 American Community Survey data, approximately 25.7 million people age five or older, 8% of the U.S. population, had limited English proficiency in 2021. The bill’s sponsors argue that language support is necessary for people accessing healthcare, housing, Medicare, disaster assistance, and other government services.
How the Bill Would Govern AI Translation and Human Review
The bill would not prohibit agencies from using AI. Instead, it would regulate the inclusion of automated tools in federal language-service workflows.
Under the full text of S. 4985, an agency could not fully replace its qualified language-assistance services with AI or machine translation. Any automated translation, transcription, or interpretation would require verification by a qualified human translator or interpreter.
In operational terms, the bill does not treat AI alone as the problem. It treats unverified and undocumented AI use as the governance problem. A translation could appear accurate and still leave an agency exposed if the agency cannot determine whether machine translation was used, who reviewed the output, or whether that reviewer was qualified.
Human review would need to address cultural context, idiomatic accuracy, specialized terminology, and clarity, not grammar alone.
The bill defines a qualified translator or interpreter as someone capable of delivering accurate, effective, and impartial communication between spoken, written, or signed languages. That person must understand relevant specialized vocabulary and follow applicable requirements for confidentiality, ethics, and impartiality.
That definition would have practical procurement consequences. Agencies would have to certify compliance with their Language Access Technical Standards annually, while AI-assisted output would require qualified human verification. Taken together, those provisions mean that describing a linguist as “qualified” may not be enough. Buyers may need records demonstrating why that person was qualified for the assignment.
This is standard documentation that a properly managed language-service provider should already retain. It includes proficiency assessments and relevant credentials for the linguists assigned to each project, evidence of appropriate subject-matter experience in legal, medical, technical or disaster-response settings, and signed confidentiality, ethics and impartiality acknowledgements.
S. 4985 does not prescribe this exact set of records. However, agencies could rely on this evidence to support their certifications, disclosures and audit responses. If enacted, the bill would make the documentation chain more consequential: an agency could substantiate its compliance only to the extent that its vendors could produce reliable records.
These are practical procurement controls rather than a verbatim document checklist contained in S. 4985. However, they are the kinds of evidence an agency may need to support its certifications, disclosures, and audit responses.
Why the Bilingual-Staff Provision Matters
The proposal would allow agencies to use demonstrably bilingual employees as an alternative to outside translators or interpreters, but only when those employees have been assessed and are qualified to provide accurate and effective communication.
This is a significant provision because conversational fluency is not the same as interpreting competence. Bilingual employees are often asked to interpret because they are immediately available, even when no one has documented their proficiency, knowledge of specialized terminology, or understanding of confidentiality and impartiality requirements.
If enacted, the bill would make that informal practice more difficult to defend. An agency relying on bilingual staff would need a credible assessment process and records showing that the employee was qualified for the communication being handled. Availability alone would not establish qualification.
Disclosure, Error Rates and Audit Trails
Agencies using AI-assisted language services would have to disclose the following information annually on LEP.gov:
- Data sources
- Known limitations
- Confidence levels
- Error rates
Their systems would also require testing for discrimination connected to language, culture, ethnicity, or another protected characteristic. Agencies would need to continuously monitor performance, investigate reported inaccuracies, and document corrective action.
For procurement teams, this creates a supply-chain documentation issue. An agency may struggle to make credible disclosures about an AI-assisted language service if its provider cannot identify the technology used, explain its known limitations, supply relevant performance information, or connect the output to a qualified human reviewer.
Each agency’s Inspector General would audit AI-assisted language systems at least once every two years to assess accuracy, fairness, cultural relevance and compliance. NIST would provide technical expertise, validation protocols and standardization tools.
The practical test is whether an agency could reconstruct a completed project months after delivery: which content was machine-assisted, which system produced it, whether that use was authorized, who reviewed the output and what qualifications that reviewer held, what quality checks were performed, and how any reported errors were resolved. A workflow that cannot answer those questions retrospectively would be difficult to defend in an audit, regardless of how well the translation reads.
S. 4985 does not expressly require project-by-project written approval before machine translation is used. Nevertheless, federal buyers can make written authorization a contractual control. It establishes that machine assistance was not introduced without the buyer’s knowledge and creates evidence that can be retained with the project record.
Language-Access Plans, Technical Standards and Public Accountability
Within one year of enactment, federal agencies would have to ensure that people with limited English proficiency could meaningfully access their public-facing programs and activities.
The bill defines meaningful access as communication that is accurate, timely, effective, available at no cost to the individual, and comparable to the access available to English-proficient users.
If enacted, the bill would require federal agencies to translate vital documents and public-facing content into languages they frequently encounter and dominant languages spoken in the United States, based on current Census Bureau data.
They would also have to offer oral interpretation, sight translation, telephone interpreting, and remote interpreting.
A vital document could include:
- An application for a government benefit or service
- A consent or complaint form
- A notice explaining rights or responsibilities
- A communication that requires a response
- Information necessary to obtain aid, training or another public service
Each agency would create a formal language-access plan covering the populations it serves, emerging language needs, service-delivery methods, emergency communications, quality measures, complaints and corrective actions.
Proposed plans would receive a 60-day public-comment period before being finalized and submitted to the Attorney General and Congress. The plans would then be published through LEP.gov.
The legislation would also establish technical standards covering in-person, telephone, and virtual interactions; paper and digital documents; websites; portals; mobile applications; and AI-assisted language services.
A central complaint system would allow individuals to report barriers to federal services. An agency would have 60 days to respond after receiving the complaint from the Attorney General.
What Federal Buyers and CCI Group Clients Should Prepare to Document
The bill directly governs federal agencies, not every hospital, university, nonprofit, court or federal contractor. However, its proposed standards could influence the organizations, vendors and public institutions that deliver services on behalf of those agencies.
Separate from S. 4985, organizations receiving federal financial assistance may have existing responsibilities under Title VI, the Rehabilitation Act, Section 1557, sector-specific regulations, grant conditions or other laws. The applicable obligations depend on the organization’s activities, funding arrangements, jurisdiction and current agency requirements.
If enacted, S. 4985 could influence how federal agencies evaluate language-service and AI vendors because agencies would remain responsible for meeting the proposed verification, qualification, disclosure, monitoring and audit requirements. A provider’s ability to produce documentation could therefore become as important as its ability to deliver the language service itself.
For buyers reviewing a language-services provider, the central questions should include:
- Can the provider identify every project in which machine translation or another automated tool was used and name the specific system?
- Can it produce written approval showing that the buyer authorized that workflow?
- Can it identify the qualified linguist who performed the human review?
- Does it retain qualification records for the linguists assigned to the work?
- Can it demonstrate relevant subject-matter experience for specialized assignments?
- Are confidentiality, ethics and impartiality records retained?
- Can it provide a review history, quality-check record, error log and corrective-action trail?
CCI Group uses machine translation and post-editing only when the client has approved that workflow in writing. Buyers should be able to determine which content was machine-assisted, who approved its use in writing, and which qualified linguist performed the human review.
An agency that does not know whether machine translation touched its public-facing content, or cannot identify the person who reviewed it, has an auditability gap even when the final output appears accurate.
Five Key Proposed Deadlines
These timelines would begin only if the bill becomes law:
| Requirement | Proposed timing |
| Meaningful access, agency plans and technical standards | Within one year of enactment |
| Public comment on proposed agency plans | 60 days |
| Compliance certification and AI disclosures | Annually |
| Inspector General audits of AI language systems | At least once every two years |
| Agency response after receiving the complaint from the Attorney General | Within 60 days |
What Happens Next?
S. 4985 must advance through the Senate Committee on Homeland Security and Governmental Affairs before receiving broader Senate consideration. H.R. 7223 must separately advance through the House Committee on Oversight and Government Reform.
The House and Senate would then need to approve identical legislative language before sending the bill to the president.
For federal buyers, the proposal is an opportunity to examine language-service records before new rules take effect. The immediate question is not whether a provider claims to use qualified professionals; it is whether the provider can produce the evidence behind that claim.
CCI Group supports government agencies and public institutions through professional translation services and interpreting services. Eligible public-sector organizations can also access these services through Sourcewell Contract #081225-IDY, a competitively solicited contract covering translation, interpretation, sign language, and CART services.
Federal procurement and language-access teams can contact CCI Group to discuss an operational readiness review focused on machine-translation disclosure, written workflow authorization, human-verification records, linguist qualifications, and audit-trail documentation.
This article provides general information about proposed legislation and does not constitute legal advice.