Connecticut SB 5 Compliance Checklist: What Every Professional Services Firm Must Do Before October 1

Connecticut's AI Responsibility and Transparency Act (SB 5) was signed by Governor Ned Lamont on May 11, 2026. The first major compliance deadline — October 1, 2026 — applies to any employer using AI as a substantial factor in employment decisions affecting Connecticut employees or applicants. That includes law firms, accounting firms, staffing agencies, and consulting practices.

This is not an overview post. If you want the background on what the law does and who it covers, read our Connecticut SB 5 overview for professional services firms. This is the implementation guide — the six concrete steps your firm must take before October 1, with firm-type callouts for each one.

---

The Short Version

Who must comply: Any employer using AI as a substantial factor in hiring, scheduling, promotion, discipline, or discharge of Connecticut employees or applicants — regardless of where the firm is headquartered.

By when: October 1, 2026 (employment AI notification and human review rights). Broader deployer obligations take effect October 1, 2027.

Three core obligations:

1. Notify affected individuals when AI was used in an employment decision — before or at the time of the decision
2. Offer human review when an AI-influenced decision produces an adverse outcome
3. Document every AI tool involved, what data it processed, and when notifications were sent

If you cannot complete those three things before October 1, 2026, the enforcement body is the Connecticut Attorney General. There is no private right of action for the employment AI provisions — but that does not mean there is no risk.

---

Step 1 — Audit Your AI Tools for Employment Decision Involvement

The threshold question for every software subscription your firm uses is: does this tool use AI, and does that AI influence any decision about an employee or applicant?

"Substantial factor" is the legal standard. AI does not have to be the only factor — it just has to meaningfully influence the outcome. A tool that generates a ranked list of candidates for a recruiter to review is almost certainly a substantial factor. A grammar-check plugin on a job posting is almost certainly not.

Tools that commonly trigger coverage:

• Applicant tracking systems (ATS) with AI scoring or ranking features — Greenhouse, Lever, Workday, iCIMS, Bullhorn
• AI-assisted resume screening tools — HireVue, Pymetrics, Applied
• Scheduling AI that manages shift assignments or staffing allocation
• Performance review platforms with AI-generated summaries or scoring
• Video interview analysis tools that rate candidate responses
• Workforce management systems with predictive scheduling

Tools that do NOT typically trigger coverage:

• Legal research AI — Westlaw Edge, Lexis+ AI, Harvey
• Tax preparation and audit AI — CCH Axcess, Thomson Reuters AI tools
• Document drafting assistants — Contract review AI, brief drafters
• General-purpose AI writing tools (ChatGPT, Claude) used for internal content
• Client-facing chatbots (covered separately under SB 5's chatbot disclosure provisions)

Your deliverable from this step: A written inventory of every software tool your firm uses, with a column for "AI in employment decisions: yes / no / unknown." Unknown requires a vendor inquiry before October 1, 2026. This document is the foundation of your compliance posture — and the first thing an enforcement inquiry will ask for.

Firm-type callouts:

• Staffing agencies: Every ATS, matching algorithm, or candidate ranking tool is your primary focus. These are almost certainly substantial-factor AI.
• Law firms: Focus on your recruiting software, associate performance platforms, and any paralegal or support staff scheduling tools. Your legal research AI is not covered.
• Accounting and consulting firms: Check your hiring workflow tools and any project staffing systems. Tax and audit AI is outside scope.

---

Step 2 — Map Which Employees and Applicants Are Covered

Once you know which tools are covered, you need to know which people trigger the notification obligation.

Covered employees: Anyone currently employed at your firm who works in Connecticut — whether they are in a Connecticut office, remote from Connecticut, or travel-based with a Connecticut home state.

Covered applicants: Anyone applying for a Connecticut position, whether they currently live in Connecticut or not. If the job is in Connecticut, the applicant is covered from the moment AI is used in screening.

Out-of-state firms with Connecticut employees: A firm headquartered in Chicago or Toronto with two Connecticut-based employees is fully covered for decisions affecting those employees. The law does not require a Connecticut office — it requires Connecticut employment.

Special note for staffing agencies: Coverage extends to placed workers and candidates — not just the agency's own internal employees. If your candidate screening process uses AI for any Connecticut placement, every candidate routed through that process has a potential notification obligation. Depending on your volume, this could mean hundreds of notification obligations per month starting October 1, 2026.

Practical mapping exercise: Pull your HR system data filtered by state. Identify every employee coded as Connecticut. Then identify every open position posted for Connecticut locations. Those are your notification populations. If your HR system can generate this list automatically, document the export process — you will need to run it on a rolling basis.

---

Step 3 — Update Your Notification Procedures

This is the step where compliance becomes operational. The law requires that notification happen before or at the time of the AI-influenced decision — not after the fact.

What the notification must include:

• A statement that AI was used in the decision
• The type of AI system used and what categories of data it processed (resume content, assessment scores, scheduling patterns, performance metrics)
• The purpose for which the AI was used
• How the individual can request human review if the outcome was adverse

Delivery method: No specific format is required. Email, a notice in your applicant portal, a letter with an offer or rejection, or a section in a candidate agreement all qualify. What matters is that it is delivered at the right time.

Template language for hiring decisions (not legal advice — have your counsel review):

This decision was informed in part by an AI-assisted evaluation tool. The tool processed [resume content / assessment responses / skills profile data] to [rank candidates / assess qualifications / evaluate fit for this role]. If you believe this decision produced an adverse outcome and would like a human review, please contact [name/email] within [timeframe]. You may also request information about the specific categories of data used.

Template language for scheduling/performance notifications:

Your [schedule assignment / performance evaluation] was informed by an AI-assisted tool that processed [scheduling data / productivity metrics / attendance records]. If you have questions about this decision or would like to request a review, please contact [name/email].

Operational implementation: Build these disclosures into your existing communication templates now — before October 1, 2026. For most firms, this is a one-afternoon update to offer letter templates, rejection email templates, and performance communication formats. The notification language does not require a lawyer to draft. It requires specificity about which tool and what data — which is why the Step 1 audit matters.

---

Step 4 — Build Your Human Review Process

When an AI-influenced decision produces an adverse outcome — a rejected application, a non-promotion, a disciplinary action, a termination — Connecticut SB 5 requires you to offer a meaningful opportunity for human review.

What "meaningful" means: The review must be conducted by a human who can actually change the outcome. A rubber-stamp process where a human confirms the AI result without independent evaluation does not satisfy the standard. The reviewer should have access to the same information the AI processed and the authority to override the result.

Who can conduct the review: The law does not require an attorney, HR specialist, or outside consultant. A senior manager, a partner, or the firm owner can conduct the human review — as long as they have the authority and information to make a different decision if warranted.

Define "adverse outcome" for your firm:

• Hiring: rejection of an applicant for a Connecticut position
• Employment: demotion, non-promotion, corrective action, performance improvement plan
• Scheduling: removal from a shift, reduction in hours for a Connecticut employee
• Termination: AI-assisted or AI-flagged termination decision

What to document: When a human review is requested, document that it was offered, who conducted it, what information they reviewed, and the outcome. If the human review changes the AI result, document that too. This documentation is your compliance record.

For staffing agencies: Your human review process needs to be scalable. If you are making hundreds of AI-influenced placement decisions per month, you cannot field individual requests ad hoc. Build a clear process — a dedicated email address, a request form, a review turnaround SLA — before October 1, 2026. Document that candidates were informed of this process in their onboarding agreements.

---

Step 5 — Update Vendor Contracts and Confirmations

Your firm's compliance is only as solid as your understanding of what your vendors' tools actually do. Many firms will discover that their ATS or HR platform uses AI in ways the firm did not realize — because the feature was marketed as a convenience, not disclosed as a decision-making system.

Request written confirmation from every AI vendor that their tool does (or does not) use AI in ways that qualify as substantial-factor employment decisions. Most enterprise HR vendors will have a compliance response ready by summer 2026 — but you need to ask, and you need to document the answer.

What to ask vendors:

• Does your product use AI to score, rank, filter, or flag candidates or employees?
• What categories of data does that AI process?
• Will you notify us if the model changes in ways that affect employment decisions?
• Do you provide compliant disclosure language your clients can use in notifications?

Update new vendor contracts: Any new AI tool procurement for HR, recruiting, or workforce management should include a clause confirming the vendor's awareness of Connecticut SB 5 and their obligations under it. This is standard risk management — if your vendor's tool creates a compliance obligation for you, the vendor should be a partner in managing it.

Watch for model updates: Connecticut SB 5's obligations are not a one-time checkbox. If your vendor updates their AI model — changing how it scores candidates, what data it uses, or how it weighs factors — that may change your notification obligations. Build a process for staying current on material vendor changes.

---

Step 6 — Document Everything

Connecticut SB 5 does not specify a documentation retention period for employment AI records. Default to your standard HR record retention policy — typically three to seven years for hiring-related records, consistent with EEOC and state employment law requirements.

What to document for each AI-influenced decision:

• Tool name and version (or date of use, if versioning is not available)
• Decision type: hiring, scheduling, performance evaluation, discipline, termination
• Employee or applicant affected (by name or ID — consistent with your HR recordkeeping)
• Notification sent: date, method, content
• Human review offered: yes/no
• If yes: who reviewed, date, outcome

Where to store it: Your existing HR file system works. A shared folder with a consistent naming convention. A row in a compliance tracking spreadsheet. What matters is that the record exists, that it is organized enough to retrieve, and that it is retained for the full retention period.

Why documentation is your first line of defense: The Connecticut Attorney General enforces SB 5. If your firm is ever the subject of an enforcement inquiry — triggered by a complaint, an audit, or an investigation — your documentation record is the first thing that will be reviewed. Firms that cannot produce it face a much worse enforcement outcome than firms that can show a disciplined compliance program.

---

Connecticut SB 5 Compliance by Firm Type

Law Firms

What is covered: Your own hiring workflows — lateral associate and attorney hiring, staff positions, paralegal scheduling. If your ATS uses AI to score or rank candidates for Connecticut positions, you are covered. If you use AI-powered associate performance review tools or productivity scoring platforms that affect Connecticut-based attorneys, those are covered.

What is not covered: Legal work product AI — Westlaw AI, Lexis+ AI, Harvey, contract review tools, brief drafters, discovery review platforms. The law targets employment decisions, not client services.

Risk level: Low to medium for most law firms. Unless you are using sophisticated AI-assisted associate review systems, your primary exposure is in recruiting.

Action items before October 1, 2026:

1. Ask your HR/recruiting software vendor whether their tool uses AI in candidate scoring or ranking
2. Update offer and rejection letter templates to include AI disclosure when applicable
3. Document a human review process for applicant requests
4. If you have an AI chatbot on your website, confirm it includes the required chatbot disclosure (a separate SB 5 obligation)

Accounting and CPA Firms

What is covered: Hiring workflows using AI — applicant tracking with AI scoring, AI-assisted scheduling for Connecticut staff, any performance management platform with AI-generated reviews or scoring.

What is not covered: Tax preparation AI, audit AI, workflow automation tools for client work. CCH Axcess, Thomson Reuters Tax & Accounting tools, Karbon's client work features — these are outside scope for the employment AI provisions.

Risk level: Low to medium. Most accounting firms' greatest exposure is in their ATS if it has AI ranking enabled, and in any AI-assisted scheduling tools for CT staff.

Action items before October 1, 2026:

1. Check your ATS (Workday, Rippling, Gusto, BambooHR, etc.) for AI-assisted candidate screening features — these are often enabled by default
2. Review scheduling tools for Connecticut office staff
3. Update hiring communications templates
4. Consider this a consulting opportunity: your Connecticut business clients have the same compliance questions, and you can walk in with a clear answer

Staffing and Recruiting Agencies

Highest-risk category. If your firm places candidates for Connecticut clients, AI-assisted screening is almost certainly a substantial factor in those placements. Candidate ranking, matching algorithms, resume scoring, video interview analysis — each one triggers notification obligations for every Connecticut candidate affected.

Risk level: High. For staffing agencies with CT candidate flow, October 1, 2026 is a hard operational deadline, not a compliance formality.

Action items before October 1, 2026:

1. Audit every vendor in your candidate workflow for AI use — Bullhorn, Avionté, JobDiva, LinkedIn Recruiter AI features, video interview tools, skills assessment platforms
2. Update candidate onboarding agreements and communication templates immediately — not in August
3. Build and document your human review process before volume arrives
4. Update your client agreements to reflect your SB 5 compliance procedures — clients placing candidates in Connecticut should know your process

For staffing-specific detail, see Connecticut SB 5 AI hiring notification law: what staffing firms must do.

Consulting Firms

What is covered: Your own internal hiring and staffing processes. Project staffing algorithms that assign consultants to engagements, skills-matching tools for internal resource allocation, and any AI-assisted performance review system affecting Connecticut-based consultants.

Risk level: Low to medium for most consulting firms, depending on how sophisticated your internal staffing tools are. Large consulting practices with AI-powered staffing allocation tools carry more exposure.

Action items before October 1, 2026:

1. Audit your internal recruiting and staffing tools for AI use
2. Update hiring and performance review communications for Connecticut staff
3. Consider the opportunity: SB 5 compliance is a new service line for consulting firms with HR or workforce advisory practices. Your Connecticut business clients have the same compliance deadline.

---

October 1, 2026 vs. October 1, 2027 — What Is Different

Connecticut SB 5 has a two-phase compliance timeline. Most professional services firms need to focus on phase one.

October 1, 2026 — Employment AI and chatbot obligations:

• AI employment decision notification requirements
• Human review rights for adverse employment outcomes
• Chatbot disclosure (AI chat interfaces must identify themselves as AI)

October 1, 2027 — Broader deployer obligations:

• Requirements for AI used in "consequential decisions" beyond employment — housing, healthcare, financial services, education
• For most professional services firms, this phase does not add significant new obligations unless you deploy AI tools that make consequential decisions for clients

Who the 2027 deadline affects: Firms using AI to make or substantially inform consequential decisions for clients — financial recommendations, credit assessments, healthcare triage. For the typical law firm, accounting firm, or staffing agency, the October 1, 2026 employment AI deadline is the operative one.

Frontier model developer requirements: Developers of large-scale foundation AI models have separate and more extensive obligations under SB 5. These do not apply to professional services firms using AI tools — they apply to the companies that build the models (Anthropic, OpenAI, Google, etc.).

---

Frequently Asked Questions

What is the October 1, 2026 deadline for Connecticut SB 5?

Any employer using AI as a substantial factor in hiring, scheduling, promotion, discipline, or discharge of Connecticut employees or applicants must begin notifying affected individuals by October 1, 2026. This is the first Connecticut AI Responsibility and Transparency Act (SB 5) compliance date.

Does Connecticut SB 5 apply if my firm is based outside Connecticut?

Yes. The law applies to any employer using AI in decisions affecting Connecticut employees or applicants — not just CT-headquartered firms. A Boston-based accounting firm with one CT-based employee is covered for decisions affecting that employee.

What counts as AI making a "substantial factor" employment decision under Connecticut SB 5?

Any AI tool that materially influences — but does not need to solely determine — a hiring, scheduling, performance, or disciplinary decision. If your ATS scores or ranks candidates using AI, that is typically a substantial factor. If a hiring manager uses AI only to draft a job posting, that is less likely to qualify.

Does my law firm need to comply with Connecticut SB 5 for AI used in client matters?

No. SB 5 regulates employment decisions, not legal work product. A firm using AI to draft motions, review contracts, or research case law is not covered. Only the firm's own hiring, performance management, and scheduling workflows are subject to the employment notification requirements.

Is there a private right of action under Connecticut SB 5?

For most provisions, no. Enforcement runs through the Connecticut Attorney General. There is limited private right of action for specific provisions (frontier model violations), but for employment AI and chatbot disclosures, enforcement is government-initiated. This does not eliminate risk — AG enforcement can still result in civil penalties.

What documentation does Connecticut SB 5 require firms to keep?

The law does not specify a retention period. Use your standard HR record retention timeline — typically three to seven years. For each AI-influenced decision, document the tool name, version or date, decision type, notification sent, and whether human review was offered and completed.

Do I need a lawyer to comply with Connecticut SB 5?

The operational steps — auditing tools, updating notification templates, building a human review process — do not require outside counsel to implement. However, if you have complex AI deployments, employment decisions affecting large Connecticut populations, or uncertainty about whether a specific tool qualifies as "substantial factor" AI, legal review is worth the investment before October 1, 2026.

---

The firms that struggle with Connecticut SB 5 will not be the ones that use AI. It will be the ones that don't know where their AI is. The law does not prohibit AI-assisted employment decisions. It requires transparency, a human review option, and a paper trail. If you can answer "which AI tools influence our hiring process, what data do they use, and how do we notify affected people," you are most of the way there. Start that audit today — not in September.

---

For multi-state context, see Multi-state AI compliance for professional services firms 2026. For the full Connecticut SB 5 overview, see Connecticut SB 5 signed: what professional services firms need to do.

The Crossing Report covers AI adoption and compliance for professional services firm owners every Monday. Subscribe free to get the weekly briefing — or go premium for implementation guides, tool comparisons, and state-by-state compliance tracking.