By Sagar Shankaran, Founder of CallSphere
Connecticut Public Act 25-113 amends the CTDPA effective July 1, 2026, adding AI/LLM disclosure, expanded sensitive-data categories, and broad minor protections. AI voice and chat get a new training-data notice obligation.
Key takeaways
Connecticut was an early state-privacy leader and now adds an explicit AI training-data disclosure requirement — small in word count, large in operational impact for any LLM-backed voice or chat agent.
The Connecticut Data Privacy Act (CTDPA) has been in force since 1 July 2023. Public Act 25-113, signed by Governor Lamont on 25 June 2025, amends the CTDPA effective 1 July 2026. The amendment expands sensitive data to include disability or treatment, status as nonbinary or transgender, neural data, financial information, and government IDs. The minor age band moves to 13–17 with a flat ban on targeted advertising and sale of minors' data, regardless of consent. Design features that significantly increase or extend a minor's use of an online service are prohibited, with explicit application to chatbots and LLMs.
The marquee AI provision: businesses that use personal data to train an AI system, particularly an LLM, must disclose that use in their privacy notice. Impact assessments must accompany profiling that produces legal or similarly significant effects and must include intended use, deployment context, benefits, risks, input categories, output description, performance metrics, known limits, transparency measures, and post-deployment monitoring. The Attorney General's enforcement report, released in 2026, signaled heightened action on minors' privacy.
Hear it before you finish reading
Talk to a live CallSphere AI voice agent in your browser — 60 seconds, no signup.
LLM-backed voice and chat agents must publish, in their privacy notice, a statement that personal data is or may be used to train the underlying model — including the categories of data, the providers, and any third-party model training. Sensitive-data processing requires consent. Minor-facing channels must implement a hard ban on targeted advertising and sale and remove engagement-extending design features. The impact assessment becomes a long-form document; copy/paste from a CCPA risk assessment will not satisfy CTDPA's enumerated elements.
CallSphere — 37 agents, 90+ tools, 115+ DB tables, 6 verticals, 50+ businesses, 4.8/5, HIPAA and SOC 2 aligned — generates the CTDPA training-data disclosure automatically from each tenant's model and provider configuration. Sensitive-data flags are first-class in the schema; consent capture is wired into every relevant intake. Minor-facing flows route to a non-engagement variant with no upsell prompts. The CTDPA impact assessment template covers all 10 enumerated elements. Pricing $149 / $499 / $1,499; 14-day trial; 22% lifetime affiliate via /affiliate; detail at /pricing and /contact.
flowchart LR
A[CT Caller] --> B[Voice Agent]
B --> C[Training-Data\nNotice]
B --> D[Sensitive Data\nConsent]
B --> E[Minor Check]
E --> F[No Targeted Ads]
B --> G[Impact Assessment]
Do we need to name the model vendor? Yes — the disclosure must name the categories of providers and the categories of data used in training.
Is fine-tuning "training"? The amendment uses broad language; treat fine-tuning as in scope unless your counsel concludes otherwise.
Still reading? Stop comparing — try CallSphere live.
CallSphere ships complete AI voice agents per industry — 14 tools for healthcare, 10 agents for real estate, 4 specialists for salons. See how it actually handles a call before you book a demo.
Does the LLM disclosure apply to inference-only deployments? Inference on a third-party model that does not train on your inputs is outside the training disclosure but inside other CTDPA obligations.
Are HIPAA-covered entities exempt? Entity-level exemption applies; verify each data flow because not all data is PHI.
What is the cure-period status? The 60-day cure period sunsetted; the AG can move directly to enforcement.
Written by
Sagar Shankaran· Founder, CallSphere
Sagar Shankaran is the founder of CallSphere, where he builds production AI voice and chat agents deployed across healthcare, hospitality, real estate, and home services. He writes about agentic AI, LLM engineering, and shipping voice agents that handle real calls in production.
See how AI voice agents work for your industry. Live demo available -- no signup required.
A founder's guide to the female voice generator landscape: AI female voices, Japanese voices, robot voices, and how CallSphere ships 57+ voices live.
MOS 4.3+ is the band where AI voice feels human. Drop below 3.6 and conversations break. Here is how to measure, improve, and alert on MOS in production AI voice using G.711, Opus, and the underlying packet loss / jitter / latency math.
How leaders should think about Claude equity research — adoption patterns, ROI, competitive dynamics, and what financial AI means for the next 12 months.
A practical engineering deep dive into Claude Sonnet 4.6 vision, covering architecture, tradeoffs, and what production teams need to know about multimodal AI.
A balanced engineering breakdown of Anthropic's Constitutional AI: what RLAIF actually does, what it cannot do, and whether it is real IP or RLHF rebranded.
Infrastructure-level look at Bedrock agents Claude, including AWS agent infrastructure, deployment topology, region availability, and cost considerations.
© 2026 CallSphere LLC. All rights reserved.