[nonimo]
EN
Download the app

Colorado AI Act: SB 26-189 replaced SB 24-205. What it asks

· Written and maintained by Nonimo

The Colorado AI Act most articles describe no longer exists in that form. SB 24-205, the 2024 law with a duty of care, risk management programs and impact assessments, was repealed and reenacted by SB 26-189, which Governor Polis signed on May 14, 2026. The replacement takes effect on January 1, 2027, and it asks for something much smaller: notice, an explanation after an adverse outcome, correction, human review and three years of records.

Two things are still open on October 13, 2026. The attorney general’s rules are a proposal with a public hearing on October 26, and a federal court order in the lawsuit xAI brought against the 2024 law pauses enforcement on terms that, read literally, also reach its replacement.

This guide goes through what the new text says, who it reaches, what an office that uses ChatGPT should take from it, and the Colorado Privacy Act, which is a different law with different people inside it.

If your question is broader than Colorado, which AI tools comply with US privacy law covers the national picture, and our map of state privacy and AI laws lists the other states.

Colorado AI Act status in October 2026: repealed, rewritten and paused

The legislature’s own bill page for SB 26-189 is titled Automated Decision-Making Technology and records the act as Chapter 131 of the 2026 session laws. Section 1 of the signed act does one thing: it repeals and reenacts, with amendments, part 17 of article 1 of title 6 of the Colorado Revised Statutes. Part 17 is where SB 24-205 lived. Same address, new text.

The new text is not the old one with dates moved. The 2024 law regulated “high risk artificial intelligence systems” and put a duty of reasonable care on developers and deployers to protect consumers from algorithmic discrimination. The 2026 law regulates “automated decision-making technology”, or ADMT, and drops the duty of care entirely. A national AI policy template written for SB 24-205 needs rewriting.

From SB 24-205 to SB 26-189 May 17, 2024 SB 24-205 signed Aug 28, 2025 delay law start moved to June 30, 2026 Apr 27, 2026 court order pauses enforcement May 14, 2026 SB 26-189 signed repeals and reenacts part 17 Oct 26, 2026 rules hearing (proposal) Jan 1, 2027 new law in force rules due the same day Filled dots have happened; open dots are still ahead on October 13, 2026
Colorado General Assembly bill pages, the signed SB 26-189 and the federal court's minute order, read October 13, 2026

What happened to SB 24-205 on the way

SB 24-205 was signed on May 17, 2024, with its main duties starting February 1, 2026. A special session bill, SB25B-004, signed August 28, 2025, pushed that start to June 30, 2026. Neither date matters now for planning, but they explain why so many compliance checklists online still lead with impact assessments and a February deadline.

There is one oddity worth knowing. Section 5 of SB 26-189 says the act takes effect January 1, 2027, except for its rulemaking pieces, which took effect on signature. Read literally, the 2024 text sat on the books from June 30, 2026 until the replacement lands. In practice it is not being enforced, for the reason in the next section.

The lawsuit and the stay of enforcement

On April 9, 2026, xAI sued Colorado’s attorney general in federal court to block SB 24-205. The Justice Department announced on April 24 that it had moved to intervene, arguing the law violates the Equal Protection Clause. The court granted intervention the same day, according to the parties’ joint filing.

On April 27, 2026, the magistrate judge granted the parties’ stipulation. The order says the attorney general shall not start enforcement, investigations included, for violations of SB 24-205 “or any legislation replacing or amending” it enacted during that session, if they occur on or before the 14th day after the court rules on xAI’s coming motion for a preliminary injunction. xAI has 28 days after the final rules to file that motion.

SB 26-189 was enacted in that session and replaces SB 24-205, so on its face the stay reaches it. Courts can change orders and this one can be revisited, so treat it as a reason the first months of 2027 may be quiet, not as permission to skip the work. The national AI policy page is the place to write down the plan either way.

What changed from SB 24-205 to SB 26-189

The quickest way to see the rewrite is side by side. The left column is what most guides still describe. The right column is the text that takes effect on January 1, 2027.

SB 24-205 (2024)SB 26-189 (2026)
What it regulatesHigh risk AI systemsCovered ADMT in consequential decisions
Core dutyReasonable care against algorithmic discriminationNotice, explanation, correction, human review
Risk program and impact assessmentRequired of deployersGone
Reporting discrimination to the AGWithin 90 days of discoveryGone
RecordsTied to assessments and reviews3 years after each decision
EnforcementAttorney general onlyAttorney general only, cure period until 2030

SB 24-205 bill summary and SB 26-189 signed act, Colorado General Assembly, read October 13, 2026.

The cut matters most for small and midsize businesses. Under the 2024 text a company using a hiring tool was looking at a written risk management program, an annual impact assessment and a public statement. Under the 2026 text the same company owes a notice, a letter after a rejection and a way to ask for review. For businesses that sell into other markets, our EU AI Act guide covers the rules that still look like the old Colorado model.

What the new law kept from the old one

Three pieces survived in a new shape. Consumers still get notice before a covered decision, a way to correct inaccurate personal data, and a path to human review after an adverse outcome. Enforcement still runs only through the attorney general under the Colorado Consumer Protection Act, and a violation is still a deceptive trade practice under C.R.S. 6-1-105.

The disclosure to the attorney general within 90 days, which made SB 24-205 unusual, did not survive. Nor did the general duty to tell people they are talking to an AI system. That idea moved, for chatbots aimed at the public, into a separate law covered further down.

Who the Colorado AI law covers: deployers, developers and consumers

SB 26-189 sorts businesses into two roles. A deployer is a person doing business in Colorado that deploys a covered ADMT. A developer is a person doing business in Colorado that develops, sells or licenses a covered ADMT, builds a component designed to be part of one, or intentionally and substantially modifies an ADMT so that it becomes covered.

Most readers of this page are deployers, if they are anything. A real estate brokerage, an insurance agency, a CPA firm or a medical practice buys software; it does not build models. The developer duties, mostly technical documentation and notice of material updates, fall on the vendor. Our page for real estate brokerages covers the AI side of a brokerage’s paperwork.

The definition of consumer reaches staff and applicants

The law borrows its definition of consumer from the Colorado Privacy Act and then widens it. Under section 6-1-1701(4)(b), a consumer includes an employee, a job applicant who is a Colorado resident, and anyone whose access to an opportunity in Colorado is evaluated in a consequential decision by a business operating there.

That one sentence is why HR teams should read this law even though they can ignore the Privacy Act. A screening tool that ranks applicants, or a system that flags employees for discipline, falls inside. Our guide to AI in HR and employee data sets the Colorado rule next to the ones in Illinois, New York City and California.

The seven covered domains

A consequential decision is one about a consumer’s access to, eligibility for, selection for or compensation for a covered domain. Section 6-1-1701(6) lists seven:

Covered domainTypical decision
EducationEnrollment, an education opportunity
EmploymentHiring, promotion, pay, discipline
Residential real estate in ColoradoA lease or a purchase
Financial or lending servicesApproval, terms, pricing
InsuranceUnderwriting, pricing, coverage, claims
Health care servicesAccess to care
Essential government services and public benefitsEligibility and renewals

C.R.S. 6-1-1701(3) and (6), as reenacted by SB 26-189.

The definition also reaches decisions about price, cost sharing or other material terms when they are reasonably likely to limit, delay, effectively deny or fundamentally alter access. A tenant screening score that raises a deposit can count, not only one that refuses the lease.

ADMT and “materially influence”: the two definitions that decide it

Coverage turns on two terms. ADMT is any technology that processes personal data and uses computation to generate output, including predictions, recommendations, classifications, rankings or scores, used to make, guide or assist a decision about an individual. That is wide enough to include a general purpose chatbot.

What narrows it is the second term. A covered ADMT is one used to materially influence a consequential decision, and section 6-1-1701(13) defines that as an output that is a non de minimis factor in the decision and that affects its outcome, including by ranking, scoring, recommending or classifying. Incidental, trivial or clerical uses are expressly outside.

What the law takes out of ADMT

Section 6-1-1701(2)(b) takes out three groups. The first is a list of ordinary technology: antivirus, firewalls, databases, data storage, spell checking, web hosting, and spreadsheets that need human analysis and use no machine learning or large language model.

The second is the one most offices will lean on. ADMT does not include “a tool used by an individual solely to summarize, organize, translate, draft, route, or present information for human review of administrative processing.” A person asking ChatGPT to tidy a letter, translate a notice or summarize a file for their own review is using exactly that kind of tool.

The third covers technology that talks to consumers in natural language, such as a support chatbot, but only if two conditions hold. It is not contracted, marketed, configured or intended for use in a consequential decision, and it is subject to an acceptable use policy that prohibits using generated content in one.

"solely"
The drafting exception covers a tool used only to summarize, organize, translate, draft, route or present. C.R.S. 6-1-1701(2)(b)(II)

The list of what is not a consequential decision

Separately, section 6-1-1701(3)(b) lists decisions that are not consequential at all: routine scheduling, administrative routing, customer service triage, communicating decisions already made, workflow management, advertising and marketing, search, and actions where the system summarizes or presents information for human review without producing a score, ranking or recommendation that materially influences the outcome.

Fraud prevention, identity verification, cybersecurity, anti money laundering and sanctions screening are also outside. So a bank’s ID check on a new account is not the kind of decision the law is about, even if software does most of it. For the ID copies themselves, how to watermark a Colorado driver’s license is the practical guide.

Is ChatGPT covered by the Colorado AI Act?

ChatGPT, Claude or Copilot is not covered or uncovered as a product. The law looks at the use. The same chat window is outside the law when a property manager asks it to draft a maintenance notice, and potentially inside it when the same manager pastes in five rental applications and asks which applicant is the best risk.

The difference is the output. A drafted notice is the drafting exception. A ranked list of applicants for a Colorado lease is a ranking used in a covered domain, and if the manager follows it, the output has affected the outcome. That is close to the definition of materially influence, and the business would be a deployer.

The acceptable use policy condition, and what OpenAI’s says

The chatbot exception requires an acceptable use policy that prohibits generated content from being used in a consequential decision. OpenAI’s usage policies, effective October 29, 2025, list among prohibited uses the “automation of high-stakes decisions in sensitive areas without human review”, and the areas they name include education, housing, employment, credit, insurance, legal, medical and essential government services.

That wording forbids automation without human review. It does not forbid using the output as one input that a human then weighs. Whether that counts as the prohibition SB 26-189 describes is not settled by the statute, and the attorney general’s rules may address it. A business should not assume the vendor’s policy moves its own use outside the law.

Where the line usually falls in an office

What staff ask the chatbot to doLikely position under SB 26-189
Draft, summarize or translate a letterDrafting exception
Sort the inbox or route a requestNot a consequential decision
Answer a customer’s question on the websiteChatbot exception, if both conditions hold
Score or rank applicants for a job or a leaseLikely covered ADMT
Recommend approve or decline on a loan or claimLikely covered ADMT

Our reading of C.R.S. 6-1-1701(2), (3) and (13) as reenacted; not legal advice, and the rules are pending.

What sits in that window is a separate question from whether the law applies. A rental application carries names, Social Security numbers and income, and a pasted file reaches the vendor under its consumer or business terms. What ChatGPT does with what you type goes through that plan by plan, and what to remove before AI lists the details that matter most.

What a deployer must do from January 1, 2027

If a business does use a covered ADMT to materially influence a consequential decision, sections 6-1-1703 to 6-1-1705 set five obligations. They apply to decisions made on or after January 1, 2027.

  1. Give notice before the decision. A clear and conspicuous notice that a covered ADMT was or will be used in a consequential decision affecting the consumer, with instructions on getting more information. A prominent public notice at the points of interaction, or a link close to the transaction, satisfies this.
  2. Explain an adverse outcome within 30 days. A plain language description of the decision and the role the ADMT played, a simple process to ask for the tool’s name, version, developer and the types and sources of personal data used, and an explanation of the consumer’s rights.
  3. Offer correction. On request after an adverse outcome, instructions for getting and correcting factually incorrect or materially inaccurate personal data. Opinions, predictions and scores do not have to be corrected.
  4. Offer meaningful human review. An opportunity for review and reconsideration, to the extent commercially reasonable, by a trained person with authority to approve, modify or override who does not default to the system’s output.
  5. Keep records for three years. Records reasonably needed to show compliance, kept at least three years after each decision, such as version identifiers and changelogs.
30 daysto explain an adverse outcome
3 yearsto keep records after each decision
60 daysto cure after an AG notice, until 2030
C.R.S. 6-1-1703, 6-1-1704(3) and 6-1-1706(3), as reenacted by SB 26-189

The notices have to reach people with disabilities and people with limited English proficiency, and a deployer does not have to disclose trade secrets, though it must say when it withholds something. Developers owe deployers the information needed to write those notices, so the vendor contract is where most of this work starts. The engagement letter language guide shows how professional firms phrase AI notices to clients.

What the developer owes the business

From January 1, 2027, the developer of a covered ADMT must give each deployer a statement of intended and known harmful uses, the categories of training data, known limitations, and instructions for monitoring and human review. It must also notify deployers of material updates, and it keeps its own records for three years. These duties apply only where the tool was marketed, configured, contracted or licensed for consequential decisions.

That last condition is a reason to read the vendor’s marketing. A general chatbot sold for drafting carries no developer duties, so a business that repurposes it to rank applicants will get no documentation at all, and the developer escapes liability for a use it never offered.

Health care, insurers and lenders: where SB 26-189 steps aside

The law has several exemptions for businesses already under a sector rulebook, and they decide a lot for the professions that read this site.

WhoWhat SB 26-189 does
HIPAA covered entities and their business associatesSections 1701 to 1706 do not apply, except for employment decisions
Covered entities using ADMT for financial assistanceMust give the patient a plain language description and review options
Insurers subject to C.R.S. 10-3-1104.9Deemed compliant in the practice of insurance
Creditors sending ECOA or FCRA noticesThose notices can satisfy this law’s notice
Schools subject to FERPAComply through their FERPA processes
FDA regulated medical devicesOutside the law

C.R.S. 6-1-1704(6) and (9) and 6-1-1708, as reenacted by SB 26-189.

The HIPAA exemption has an employment exception

For a medical practice the HIPAA exemption is the headline, but it stops at the HR door. A clinic that uses an AI tool to screen job applicants is a deployer for that purpose like any other employer. For a health care provider the exemption also applies only if it operates from a location in Colorado, and every covered entity must give patients a general notice that it uses advanced technologies, including covered ADMT.

The exemption from this law is not an exemption from HIPAA, which still governs what goes into the tool. Our HIPAA compliant AI guide covers the business associate agreements, and PHI versus PII explains why the HR file is not PHI in the first place.

Insurance agencies and the Division of Insurance

Insurers and affiliates already subject to C.R.S. 10-3-1104.9, Colorado’s statute on insurers’ use of external consumer data, algorithms and predictive models, are treated as compliant in the practice of insurance. Section 3 of SB 26-189 lets the insurance commissioner adopt new notice rules for them. That treatment does not reach an insurer’s own employment decisions. Our page for insurance agencies covers what agents paste into AI.

Enforcement: the attorney general, a cure period and no private suits

Section 6-1-1706 makes the attorney general the only enforcer of the notice, record and consumer rights duties, through the Colorado Consumer Protection Act. A violation is a deceptive trade practice. The act creates no new private right of action, and it says so twice.

Before any enforcement action, the attorney general must send a notice of violation if a cure is possible, and the business has 60 days to fix it. Knowing or repeated violations skip that step. The cure provision is repealed on January 1, 2030, so the softer start has an end date. Our breach guide covers the separate question of what to do after a paste goes wrong.

Discrimination claims still run under existing law

Section 6-1-1707 keeps the route that matters most to plaintiffs. A developer or deployer can be liable under the Colorado Anti-Discrimination Act for a consequential decision materially influenced by covered ADMT, and fault is allocated by relative fault. A developer is liable only where its tool was used as it intended, documented or marketed.

The same section voids any contract term that would indemnify a developer or deployer for its own acts in violation of Colorado anti-discrimination law. That is unusual, and it changes vendor negotiations: a business cannot simply push all of that risk onto its software supplier in the contract, or the supplier onto the business.

The attorney general’s rules: proposed, not final

The statute requires two sets of rules by January 1, 2027: one on the post adverse outcome disclosures, which may add sector specific guidance and standards for describing the ADMT’s role, and one on the consumer rights. It also allows rules clarifying “materially influence”, including presumptions and illustrative examples.

DateStep, per the Colorado Department of Law
Through July 13, 2026Informal comments before formal rulemaking
August 11, 2026Proposed rules filed with the Secretary of State
October 6, 2026Interim draft released
October 26, 2026Public hearing in Denver; written comments close that day
January 1, 2027Statutory deadline for the required rules

coag.gov/ai, read October 13, 2026. The rules cover SB 26-189 and the Chatbot Safety Act together.

Until the rules are adopted, the details of the 30 day letter are not fixed, and the clarification of materially influence that offices most need may change the answer for borderline uses. Anything a vendor tells you the rules “require” today is a reading of a draft. If the hearing continues past October 26, the comment period continues to its last day.

The chatbot law is a different statute

The rulemaking also covers HB 26-1263, the Chatbot Safety Act, signed May 29, 2026 according to the legislature’s bill page. It sets requirements for operators of publicly available conversational AI services, including age estimation, protections for minors, telling users they are talking to AI, self harm protocols and annual reports to the attorney general, with the operator duties starting January 1, 2027.

It is aimed at companies running public chatbots, not at offices whose staff use one. A firm whose staff use ChatGPT internally is not running a publicly available service. Writing that distinction into your AI policy saves a later argument.

The Colorado Privacy Act and AI tools: a separate law with its own rules

The Colorado Privacy Act, signed as SB 21-190 on July 7, 2021, took effect July 1, 2023. It is enforced by the attorney general and district attorneys, and it is the law that governs what personal data an office hands to an AI vendor in the first place. It is not part of the AI law and it covers different people.

It applies to businesses, nonprofits included, that do business in Colorado or target Colorado residents and that either process personal data of 100,000 or more Colorado residents in a calendar year, or earn revenue or discounts from selling personal data while processing the data of 25,000 or more. The attorney general’s FAQ confirms the 100,000 means Colorado residents, not people nationwide.

Colorado Privacy ActWhat it means for AI use
Opt out of certain profilingResidents can refuse certain kinds of profiling
Consent for sensitive dataHealth, identifying biometrics, a child’s data and more
Data protection assessmentsRequired before selling data, processing sensitive data or risky profiling
Processor dutiesAn AI vendor processing for you follows your instructions
Employee dataOutside: the CPA excludes people acting as employees or applicants
Universal opt-outRecognized signals honored since July 1, 2024

Colorado Attorney General, Colorado Privacy Act page and FAQ, read October 13, 2026.

Controllers, processors and the vendor contract

When a covered business pastes customer data into an AI tool for its own purposes, it is the controller and the vendor processing on its behalf is a processor. The processor must follow the controller’s instructions, keep the data confidential and help with consumer requests, under C.R.S. 6-1-1305. A personal chatbot account, run on the vendor’s own consumer terms, is a poor fit for that relationship.

Consent for sensitive data has to be affirmative, specific and informed; accepting broad terms of service does not count. A customer’s health condition or a child’s details pasted into a tool for a new purpose is the kind of secondary use the attorney general’s FAQ says needs fresh consent. Our CPA firm page covers the client files where this comes up most.

The biometric amendment reaches employers

HB 24-1130, signed May 31, 2024 and effective July 1, 2025, added biometric protections to the Privacy Act. Controllers need a written retention and deletion policy and must meet disclosure and consent requirements before collecting a biometric identifier, and the act restricts why an employer may ask employees to consent. A face photo on an ID scan is the kind of file to keep out of a chat window. Redacting a driver’s license copy shows how.

Two laws, two different sets of people

The useful thing to remember about Colorado is that the two laws do not overlap neatly. The Privacy Act excludes employees and job applicants; the AI law includes them. The Privacy Act has volume thresholds; the AI law has none, so a ten person firm that ranks Colorado applicants with software can be a deployer. A company can be inside one, the other, both or neither.

What a Colorado business can do before January 1, 2027

Most of the work is an inventory. Here is the order that tends to save time:

  1. List where software scores, ranks or recommends in hiring, leasing, lending, insurance, health access, education or public benefits. Those are the only places SB 26-189 can apply.
  2. For each, ask the vendor whether the product is marketed for consequential decisions and request its developer documentation. If the answer is no, decide whether your use changes that.
  3. Separate drafting from deciding in your AI policy, and say who may use a chatbot to rank people. Our US policy template has the clauses.
  4. Draft the notice, the 30 day adverse outcome letter and the review process, then revisit them once the attorney general adopts final rules.
  5. Check the Privacy Act thresholds on their own, and whether any AI vendor is handling Colorado residents’ data without a processor contract.

Nonimo on a Colorado office computer

Nonimo is an app for Mac and Windows. Whoever is drafting highlights the text and presses one key, and the names, addresses, Social Security numbers and other identifiers it recognizes become placeholders on that computer, before ChatGPT, Claude or Copilot receive anything. When the answer comes back, the real details are put back on the same machine.

Typed                                         What reached the AI tool
Applicant: Delphine Okonkwo                  Applicant: [PERSON_1]
SSN: 000-12-3456                              SSN: [REFERENCE_1]
Phone: (303) 555-0147                         Phone: [PHONE_1]
Unit: 1400 Example Street, Denver, CO 80202   Unit: [ADDRESS_1]

The applicant is invented, Social Security never issues area 000, the number sits in the 555-0100 to 555-0199 block kept for fiction, and the address uses a placeholder street name.

That does not decide whether a use is covered by SB 26-189, which depends on what the tool’s output is used for.

It reduces what personal data the AI vendor holds, which is what the Privacy Act and the vendor contract care about. The apps keep each conversation encrypted on your disk; what leaves once a day is a usage count with no text, as the security page sets out. The app for US offices is the place to start, with central rollout for organizations.

Sources

All read October 13, 2026.

Common questions

Is the Colorado AI Act still in effect?

Not in its 2024 form. SB 26-189, signed May 14, 2026, repeals and reenacts the part of the Colorado statutes that SB 24-205 created, and the new version takes effect January 1, 2027 for consequential decisions made from that date. The duty of care, the risk management program and the impact assessments are gone. What remains is notice, an explanation after an adverse outcome, correction, human review and records.

When do businesses have to comply with SB 26-189?

From January 1, 2027. Section 5 of the act says it takes effect on that date and applies to consequential decisions made on or after it. Only the rulemaking provisions took effect when it was signed. The attorney general must adopt rules on the adverse outcome disclosures and the consumer rights by the same date, and those rules were still a proposal on October 13, 2026.

Does using ChatGPT at work make my business a deployer?

Not by itself. A deployer is a business that uses covered automated decision making technology to materially influence a consequential decision, such as hiring, a loan or a lease. The definition excludes a tool used by an individual only to summarize, organize, translate or draft for human review. Drafting an email is outside; ranking applicants with the same tool may not be.

What is a consequential decision under the Colorado law?

A decision about a consumer's access to, eligibility for, selection for or compensation for one of seven covered domains: education, employment, residential real estate in Colorado, financial or lending services, insurance, health care services, and essential government services. It also covers pricing and material terms that would effectively deny or fundamentally alter that access. Scheduling, routing and customer service triage are excluded.

Can a consumer sue under the Colorado AI Act?

No. SB 26-189 says twice that it creates no new private right of action, and only the attorney general enforces it, through the Colorado Consumer Protection Act. It does keep existing remedies. A person can still bring a discrimination claim under the Colorado Anti-Discrimination Act, and the new law sets how fault is shared between the developer and the business that used the tool.

Does the Colorado AI law cover employees and job applicants?

Yes, and that is a difference from the Colorado Privacy Act. SB 26-189 defines consumer to include an employee and a job applicant who is a Colorado resident. The Colorado Privacy Act excludes people acting as employees or applicants. So an AI screening tool can trigger the AI law while the HR file stays outside the privacy law.

What is the status of the attorney general's rules?

Proposed, not final. The Department of Law filed draft rules on August 11, 2026, released an interim draft on October 6, and set a public hearing for October 26, 2026 in Denver, with written comments accepted through that day. The statute requires rules on the adverse outcome disclosures and consumer rights by January 1, 2027.

How does the Colorado Privacy Act apply to AI tools?

It applies to businesses that handle personal data of 100,000 or more Colorado residents a year, or that sell data of 25,000 or more. It gives residents a right to opt out of certain profiling, requires consent before processing sensitive data, and requires data protection assessments for risky processing. An AI vendor that processes data on your behalf is a processor and must follow your instructions.

Does Nonimo make a business compliant with SB 26-189?

No tool does that, because the law is about how a decision is made and explained. What Nonimo does is narrower: on the computer where the text is typed, it swaps names and identifiers for placeholders before the text reaches ChatGPT, Claude or Copilot, and puts them back in the reply. That keeps less personal data with the AI vendor.

Tools

Remove personal information from a file or text before AI reads it. Or cover the details on a copy of your ID and add a watermark before you send it.

All tools →