AI Chatbots and Wiretapping Laws: When Your Customer Service Bot Triggers CIPA and the Federal Wiretap Act

If your business uses a chat widget on its website — the kind that pops up in the corner and asks “How can I help you today?” — you may already be sitting inside an active litigation target. Over the past four years, plaintiffs’ attorneys have filed hundreds of class action lawsuits under California’s wiretapping statute and, increasingly, under analogous laws in other states, arguing that deploying a third-party AI chatbot constitutes illegal interception of customer communications. The damages available under these statutes run to $5,000 per violation, and class actions can sweep in every customer who visited your site from a covered state.

This is not a theoretical risk. Courts have let these cases proceed. The litigation wave is expanding beyond California. And the legal framework that governs it — a patchwork of state wiretapping laws layered on top of a 1968 federal statute — is not intuitive. Business owners who deploy AI customer service tools deserve a plain-language explanation of what the law actually says, where courts have landed, and what practical steps reduce exposure.

What CIPA Actually Says — and Why It Reaches Your Website

The California Invasion of Privacy Act was enacted in 1967 as a criminal wiretapping statute. The California Legislature declared in the statute’s preamble that the right to privacy is “a fundamental and compelling interest” and that “advances in science and technology have led to the development of new devices and techniques for the surveillance of individuals.” That framing was prescient, even if the legislature could not have imagined AI-powered chat assistants.

Section 631: The Wiretapping Prohibition

California Penal Code § 631(a) makes it unlawful for any person to intentionally tap or make any unauthorized connection with any telegraph or telephone wire, line, cable, or instrument. More relevantly for digital commerce, it also prohibits willfully and without the consent of all parties reading, attempting to read, or learning the contents of any message or communication while it is “in transit or passing over any wire, line, or cable” within California — and it extends liability to anyone who aids, agrees with, employs, or conspires with any person to do the same.

That last clause — the aiding and abetting provision — is what makes this statute so potent for website operators. Even if the website owner is considered a “party” to the conversation and therefore cannot directly wiretap its own communications, it can still be held secondarily liable for enabling a third-party vendor to intercept those communications without the customer’s prior consent.

Section 632.7: The All-Party Consent Rule for Phone Communications

Section 632.7 addresses a different context: communications transmitted between cellular phones, cordless phones, or between those devices and landlines. This section requires the consent of all parties before anyone intercepts or records such a communication. Unlike earlier provisions, section 632.7 does not require a showing of malicious intent, and California courts have interpreted it to reach telephone communications regardless of whether the parties expected confidentiality. For businesses using AI voice assistants to handle inbound customer phone calls — an increasingly common deployment pattern — this provision is the one most likely to trigger liability.

The Civil Damages Provision: Section 637.2

CIPA is a criminal statute but it also provides a robust private right of action. California Penal Code § 637.2 authorizes any person whose communication has been intercepted in violation of CIPA to sue for the greater of $5,000 per violation or three times their actual damages. The California Legislature amended the statute in 2017 to make clear that the $5,000 figure applies per violation, not per lawsuit. Critically, the statute explicitly provides that a plaintiff does not need to prove actual injury to recover statutory damages. In a class action covering millions of website visitors, the math becomes catastrophic quickly.

The Litigation Wave: How Plaintiffs’ Lawyers Are Deploying These Statutes

The first significant wave of CIPA litigation targeting website technology began around 2021 and 2022, initially focused on “session replay” software — tools that record a visitor’s keystrokes, mouse movements, and page interactions in real time. Plaintiffs argued that installing this software on a website amounted to wiretapping because a third-party vendor was intercepting the visitor’s communications as they happened, not after the fact.

Javier v. Assurance IQ: The Ninth Circuit Opens the Door

The case that put this litigation wave on the map was Javier v. Assurance IQ, LLC, which the Ninth Circuit addressed in a May 2022 decision. The plaintiff visited an insurance-quoting website and filled out a questionnaire. A third-party product called TrustedForm, provided by ActiveProspect, ran on the website and recorded the plaintiff’s keystrokes and mouse clicks. The plaintiff was not informed that this recording was happening until after he had already completed the form, at which point he was presented with the site’s privacy policy.

The district court initially dismissed the case on the ground that the plaintiff had retroactively consented to the recording by agreeing to the privacy policy. The Ninth Circuit reversed that ruling. The court held that section 631 requires the prior consent of all parties to a communication — consent that comes after the fact is insufficient. That ruling produced a wave of more than 75 class action lawsuits in 2022 alone targeting retailers, insurers, and other companies using chat features on their websites.

The ConverseNow Case: AI Voice Bots and the Phone Call Problem

The most significant recent development came in August 2025, when the United States District Court for the Northern District of California denied a motion to dismiss a CIPA class action against ConverseNow Technologies. ConverseNow provides AI-powered voice assistants that restaurants use to handle inbound customer phone calls. In January 2025, a plaintiff named Eliza Taylor sued ConverseNow, alleging that when she called a Domino’s Pizza location, her call was intercepted and routed through ConverseNow’s servers without her knowledge or consent.

The court applied the “capability” approach to determine whether ConverseNow was a third-party interceptor. Under this approach, a vendor is considered a third party when it has the capability to use intercepted communications for purposes independent of the business that deployed it. Because ConverseNow was using the recorded conversations for its own AI training and commercial development, the court found it could not claim the protection of the “party to conversation” defense.

The “Party to the Conversation” Defense — and Why It Is Not a Safe Harbor

The most commonly cited defense in CIPA chatbot litigation is the “party exception.” California courts have consistently held that a party to a communication cannot be held liable for intercepting its own communication. For the website operator itself, this generally means the business is not directly liable under section 631 for receiving messages its own customers send to it.

But the party exception does not reach as far as many businesses assume, for two distinct reasons.

The Aiding and Abetting Problem

Even if the website operator is a party to the conversation, the operator can still be held liable for aiding and abetting a third-party vendor that is not a party. Under CIPA’s express text, liability extends to anyone who “aids, agrees with, employs, or conspires with any person to do or permit any act prohibited” by the statute. Courts have allowed aiding and abetting claims against website operators to proceed even when direct wiretapping claims fail, because the operator knowingly integrated the vendor’s code into its website, enabling the interception.

The Extension Versus Capability Split

Some courts apply an “extension” test, asking whether the vendor functions merely as an extension of the website owner. Under this approach, a chatbot vendor that does nothing with the data other than provide it back to the business might be treated as an extension of the business and covered by the party exception. Other courts apply the “capability” test: if the vendor has the ability to use the intercepted communications for its own independent commercial purposes — training its AI, improving its products, analyzing call data across clients — then the vendor is a third-party eavesdropper. Most modern AI customer service platforms are built precisely to learn from every conversation they process. That commercial value proposition is also what makes the “extension” argument unavailable when the vendor’s own materials confirm that conversation data fuels its development.

The Consent Timing Problem

Even where a vendor might otherwise qualify for the party exception, the Javier ruling imposes an independent requirement. A disclosure buried in a terms of service page that a customer agrees to only after the chatbot has already begun processing their input does not constitute the prior consent that CIPA demands. Many businesses implement chatbots that begin recording interactions the moment a user opens the chat window, with any privacy disclosures appearing later in the flow. Under the Ninth Circuit’s analysis, that sequence is inadequate.

What the Federal Wiretap Act Adds to the Picture

Alongside state law, the federal Electronic Communications Privacy Act makes it a federal crime to intentionally intercept any wire, oral, or electronic communication. The relevant provision, 18 U.S.C. § 2511, provides an independent basis for liability.

The One-Party Consent Default

The federal statute defaults to a one-party consent standard. Under 18 U.S.C. § 2511(2)(d), it is not unlawful to intercept an electronic communication if one of the parties to that communication has given prior consent — and the party giving consent can be the person doing the recording. This is a materially lower standard than California law. For businesses operating entirely outside California and other all-party consent states, the Federal Wiretap Act’s one-party consent exception provides significant breathing room. There is a critical limitation, however: the exception disappears if the recording is made for the purpose of committing any criminal or tortious act.

Federal Civil Remedies

The Federal Wiretap Act provides a private civil right of action under 18 U.S.C. § 2520. A person whose communications are illegally intercepted may recover the greater of actual damages or statutory damages of $100 per day for each day of the violation or $10,000, whichever is greater — plus punitive damages and attorney’s fees. In practice, plaintiffs overwhelmingly prefer CIPA over the Federal Wiretap Act in chatbot cases because CIPA’s $5,000 per-violation damages are higher, the all-party consent standard is stricter and easier to establish, and the aiding and abetting provision provides a direct path to holding website operators liable even when the primary interception is by a vendor.

The Multi-State Patchwork: California Is Not Alone

Businesses that serve customers across the country should understand that a number of other states have enacted analogous all-party consent wiretapping laws. Florida’s Security of Communications Act requires the consent of all parties before a wire, oral, or electronic communication may be intercepted or recorded. Florida experienced a sharp rise in class action filings between 2021 and 2024. Illinois’s Eavesdropping Act similarly requires all-party consent. Pennsylvania, Maryland, Washington, Nevada, Montana, New Hampshire, and Delaware round out the list of states with all-party consent statutes that could be applied to AI customer service deployments.

The practical consequence for multi-state businesses is that they cannot simply comply with the law of their home state and assume they are covered. Any chatbot deployment that reaches California visitors — or visitors in any other all-party consent state — must be evaluated under that state’s law as well. Establishing a single compliance standard keyed to the strictest applicable jurisdiction is the most defensible approach.

A Practical Compliance Framework for Businesses

The compliance steps this landscape demands are actually fairly concrete. Businesses that take them systematically will be in a substantially better position than those that do not — both in avoiding lawsuits and in defending them if they arise.

Audit Your Current Chatbot Deployment

Before anything else, a business needs to understand what its chatbot vendor is actually doing with conversation data. Read the vendor’s terms of service, data processing agreement, and privacy policy with specific questions in mind: Does the vendor use conversation data to train its own AI models? Does it share data with third parties? Does it use the data for its own advertising or product development? If the answers are yes, or if the vendor’s agreements are ambiguous, the business faces meaningful exposure under the “capability” standard. You cannot manage risk you have not identified.

Obtain Prior, Explicit Consent — Before the Chat Begins

The sequence of disclosure matters enormously. The prior consent requirement means the consent mechanism must be presented and accepted before the chat session begins recording. A banner that appears when the chat window opens, clearly stating that the conversation may be recorded and processed by the company and its service providers, and requiring the user to affirmatively click “I agree” before proceeding, is the appropriate structure. Passive disclosure buried in a terms of service document will not satisfy the prior consent requirement. The disclosure should specifically identify whether a third-party AI vendor is processing the conversation.

Renegotiate Your Vendor Agreements

If your chatbot vendor’s standard agreement permits it to use conversation data for its own AI training, model improvement, or product development, that provision is a liability. Businesses should seek contractual provisions that explicitly restrict the vendor to processing conversation data solely for the purpose of providing the contracted service, prohibit the vendor from using data for its own commercial purposes, require the vendor to indemnify the business for claims arising from the vendor’s independent data use, and require deletion of conversation data after a defined retention period. A vendor that refuses to limit its data use is, from a CIPA standpoint, a third-party interceptor regardless of how the contract characterizes the relationship.

Update Your Privacy Policy with Specificity

Generic privacy policies that disclose “we may share your information with service providers” are almost certainly insufficient for CIPA compliance. The privacy policy should specifically identify the categories of third-party vendors that process customer chat communications, describe the purposes for which those vendors use the data, and state clearly whether conversations are recorded and retained.

Monitor Vendor Changes Regularly

AI vendors update their systems frequently, often without prominently alerting business customers to the changes. A vendor that initially agreed not to use data for AI training may quietly introduce a new feature that does exactly that. Building into vendor management practice a regular review of vendor terms, product updates, and privacy policies — at least annually, and whenever a significant product update is announced — is an important control.

The Takeaway

Wiretapping law is an area where the law was written decades before the technology it is now being applied to existed. That mismatch creates genuine legal uncertainty — courts are still working out which analytical frameworks apply. But legal uncertainty does not mean legal safety. The case law is clear on several things: prior consent is required before interception begins; third-party vendors that use conversation data for their own purposes are not protected by the party exception; and the $5,000 per-violation statutory damages structure makes these claims attractive for class action litigation even when the underlying injury is modest.

The businesses that navigate this environment most successfully will be the ones that treat chatbot deployment as a legal compliance event, not just a technology decision. That means auditing vendor relationships before deployment, designing consent flows that are honest and clearly timed, updating vendor agreements to control secondary data use, and monitoring the rapidly evolving statutory landscape.

If you are uncertain whether your current chatbot deployment meets these standards, consulting with counsel before a demand letter arrives is considerably less expensive than responding to one after it does.

This post is for general informational purposes only and does not constitute legal advice. Businesses with questions about their specific deployments should consult qualified legal counsel.