October 1. That’s the date.
Not January 1, 2027, the AI companion deadline from the same bill that generated most of the coverage when Connecticut’s SB 5 passed. The Automated Employment Decision Technology provisions have an earlier clock, per analysis from Littler Mendelson, one of the country’s leading employment law practices. If you’re using AI to screen resumes, rank candidates, assess performance, or inform compensation decisions in Connecticut, October 1, 2026 is the date your compliance program needs to be ready, not the one that got the press coverage.
SB 5’s AEDT framework requires bias testing before deployment and on a defined ongoing basis. It requires disclosure to employees and job candidates when an automated decision tool is used in a decision affecting them. And per analysis from DLA Piper, SB 5 creates a program for third-party certification of AI models used in employment decisions, a feature legal analysts describe as among the first of its kind at the state level in the US. That third-party certification requirement isn’t just a compliance step. It’s a procurement question: which AI vendors serving the employment market will be able to produce certified models, and which won’t?
Connecticut SB 5 AEDT Readiness, Before October 1
- Identify all AI tools used in hiring, performance evaluation, or compensation decisions in Connecticut
- Confirm vendor certification status under SB 5's third-party program, or timeline to certification
- Establish bias testing cadence and documentation protocol per SB 5 requirements
- Draft candidate and employee disclosure notices for AEDT use
- Confirm statutory requirements against CT General Assembly SB 5 text, don't build solely on law firm summaries
The distinction between the two SB 5 deadlines matters for compliance architecture. Connecticut AI companion rules, covering emotional support bots, AI relationships, synthetic companionship products, take effect January 1, 2027, and they’ve driven most of the legal commentary on the bill. The AEDT employment provisions hit three months earlier and affect a far larger and more immediate set of enterprise use cases.
Most employers deploying AI hiring tools today haven’t structured their vendor relationships around bias audit requirements. They’ve purchased a product and accepted the vendor’s assurances. Connecticut’s third-party certification requirement breaks that pattern. The employer bears responsibility for ensuring the model they’re using has been certified, not just for using a certified vendor. That’s a different procurement standard, and it puts pressure on AI HR technology vendors to move through certification programs before October, or risk losing accounts in Connecticut.
Four and a half months isn’t generous for a bias testing and certification cycle, particularly if vendors haven’t initiated the process yet. The catch is that “per law firm analysis” isn’t the same as reading the statutory text yourself, and the specific certification program mechanics, which certifying bodies are authorized, what the testing methodology requires, and what “bias” is defined to mean under the statute, are details that should be confirmed against the Connecticut General Assembly’s published text of SB 5 before your compliance team builds around them.
What to Watch
The broader pattern is worth naming. Connecticut joins Colorado’s ADMT framework, California’s “No Robo Bosses” Act, and New York City’s Local Law 144 in applying specific requirements to employment AI. These frameworks don’t agree on scope, certification requirements, effective dates, or enforcement mechanisms. Multi-state employers aren’t facing one employment AI compliance problem. They’re facing four, with more likely coming.
The real question isn’t whether SB 5 applies to you. It’s whether your AI HR vendor is on track for certification before October.