Connecticut Just Barred the “Algorithm Did It” Defense
Connecticut’s new AI workplace law is not just about disclosure. It tells employers they cannot point to automated decision software as their discrimination defense.
Connecticut just changed the AI hiring playbook: employers can no longer defend a discrimination claim by saying the algorithm made the decision.
What Actually Happened
Governor Lamont signed Public Act 26-15 in May, creating a two-phase compliance framework for employers using AI in workplace decisions.
Phase 2 gets most of the compliance attention. Beginning October 1, 2027, employers will face pre-decision disclosure requirements for applicants and employees, along with bot-disclosure obligations when AI systems interact with them.
Phase 1 comes first. Beginning October 1, 2026, employers cannot use an automated employment-related decision technology as a defense to a discrimination complaint.
Connecticut calls this “automated employment-related decision technology.” In plain English: software that scores, ranks, recommends, screens, classifies, or otherwise materially influences an employment decision.
So the key question is not just: did you disclose the tool?
It is: when the tool becomes part of a contested employment decision, what evidence do you actually have?
Why It Matters
The plaintiffs’ bar reads this one way: employers cannot hide behind “the machine made the call, not us.” The algorithm-as-neutral-arbiter argument — the one that tries to break the causal chain between human bias and adverse employment outcomes — just got much weaker, at least in Connecticut.
The defense bar reads it another way: plaintiffs now have a statutory answer when an employer points to software as the neutral decisionmaker.
Both readings point to the same practical result: automation is not insulation.
Connecticut did not make AI hiring tools illegal. And it did not make every AI-assisted employment decision discriminatory.
But it did legislate away the easy version of the defense.
The employer cannot end the conversation by saying, “we used neutral software.” The defense that remains is what lawyers call “fact-intensive”: what did the tool do, how was it tested, what did the testing show, and what did the employer do with those results?
That creates two immediate action items.
Vendor contracts. Connecticut makes the employer-facing risk harder to outsource. Contracts should not merely say the vendor provides “objective” software. They should address testing, audit support, documentation, cooperation in claims, indemnity, and responsibility for statutory notice obligations.
Business justification documentation. Document facts to show: the tool was used for legitimate business reasons, produced non-discriminatory outcomes, and here is the evidence. That requires documentation most companies are not building. The time to start is before the claim, not after.
The Point
Connecticut’s new law bars “the algorithm decided, not us” as a defense to discrimination claims.
Automation is not the defense. Testing, documentation, and business justification are the defense.
Go Deeper: Full briefing on AI State Legislation — recent developments, sourced: lawsnap.com/tracker/legal-intelligence/tag/ai-state-legislation/
On Our Radar
Corporate: Trump signed an AI executive order on June 2 establishing a voluntary 30-day pre-release review window for frontier models — and explicitly disclaiming mandatory licensing, permitting, or preclearance. Companies can decline without legal consequence; for boards asking what the federal government now requires of AI developers, the answer is nothing binding. White House Issues Executive Order Targeting Frontier AI Models
Tech: The EU AI Act trilogue deal is done — a May 13 provisional agreement between the Parliament, Council, and Commission postponed key high-risk AI compliance deadlines that were set to hit August 2026. Companies building toward that window need to know what was deferred and what wasn’t before slowing remediation work. Digital Omnibus on AI: The Trilogue Deal Is Done
Litigator: Courts keep sanctioning lawyers for AI-generated hallucinations even when AI use is disclosed upfront — because disclosure tells the reader an AI was used, not that the citations were verified. The standard emerging from sanctions decisions is not “did you flag it” but “did you check it.” Why AI Disclosure Rules Do Not Stop Hallucinated Citations
Each of these feeds updates daily. Start at lawsnap.com/tracker/legal-intelligence — topic clusters with briefings, or go straight to your practice area: Corporate · Litigator · Tech · In-house