What Does the Ninth Circuit Ruling on AI Agents Mean for My Business's Liability When Using Voice AI for Outbound Calls?

What does the Ninth Circuit ruling on AI agents mean for my business's liability when using voice AI for outbound calls? Howard v. Republican National Committee, decided January 13, 2026, narrows one technical question about optional video attachments, not voice-call liability generally. Liability stays with the business, and TCPA exposure of at least $500 per violation remains intact.

Does the Ninth Circuit ruling create a broad exemption for AI voice calls?

No, the Ninth Circuit ruling does not exempt AI voice calls from TCPA liability. Howard v. Republican National Committee, decided January 13, 2026, addressed only whether a text message with an attached, optional video and prerecorded audio counts as a call 'made or initiated' with an artificial voice, and the court said it does not.

The case turned on a narrow fact pattern: a text message linked to an optional video, and because the recipient had to press play to hear the audio, the Ninth Circuit found the message was not initiated with an artificial or prerecorded voice. According to Orrick's infobytes legal update on the decision, the panel held the TCPA "does not prohibit texts with optional video content" under that specific setup. The holding turns entirely on how the communication started, not on whether AI touched any part of the chain. Text messages themselves remain TCPA 'calls' in the Ninth Circuit, a point the Consumer Financial Services Law Monitor's January 2026 analysis of the same case underscores for any business weighing a channel switch.

How does the ruling affect liability for businesses using AI voice agents?

Liability for AI voice agents is largely unaffected by the Ninth Circuit's ruling. An outbound AI agent that automatically dials a number and begins speaking with a synthetic voice at the start of the call still uses an artificial voice at initiation, the exact scenario the TCPA was written to cover, regardless of how a separate text-and-video case was decided.

The FCC has stated that AI-generated and cloned voices fall within the TCPA's 'artificial or prerecorded voice' category, generally requiring prior express consent absent an applicable exception. That interpretation sits outside the scope of Howard, which never reached the question of whether a live, automated AI voice counts as artificial. A business running outbound AI calling for appointment reminders, collections, or lead qualification is still squarely inside the TCPA's consent framework. An enterprise voice AI platform, the kind Agxntsix deploys for inbound and outbound call automation, has to be built around that reality rather than around a hope that some future ruling narrows the definition further.

What are the operational implications of the ruling for enterprise voice AI?

The ruling's operational implication is narrow: it helps separate interactive, user-triggered content from calls that automatically begin with prerecorded or synthetic speech. It gives businesses a clearer line for classifying click-to-play video attachments as lower risk, but it does not change how an outbound AI voice program, consent capture, or do-not-call suppression must be designed.

Picture a dental group layering an optional recorded video into an appointment-confirmation text versus the same group running an outbound AI agent that calls a patient list and starts talking the moment the line connects: the first increasingly sits in lower-risk territory inside the Ninth Circuit, the second does not move at all. The practical task for an operations leader is pre-deployment classification, mapping every channel, voice call, text, text-with-video, to how it initiates, who answers, and what consent record supports it before any campaign goes live. That classification step is also where embedded AI consulting earns its fee: the business, not the vendor, controls campaign purpose, audience, and the record of whether a number was ever validly consented.

How wide is the gap between AI adoption and AI governance controls?

The gap between AI adoption and AI governance is wide and measurable. A 2025 EY survey of 975 C-suite leaders across 21 countries found that 72% of organizations had integrated and scaled AI across most or all initiatives, far outpacing the governance controls needed to manage the resulting liability.

That same EY survey found only about one-third of organizations had controls covering every element of its responsible-AI framework, with strong controls in just three of nine responsible-AI areas on average, including accountability, compliance, and security. Separately, 76% of companies in that 2025 EY survey were using or planning to use agentic AI within a year, while only 56% of executives said they were familiar with its associated risks. A later 2025 EY survey found that 99% of organizations reported financial losses from AI-related risks, with 64% reporting losses above $1 million and an average reported loss of $4.4 million; 57% cited noncompliance with AI regulations as a major risk. ISG's 2025 enterprise-AI study reported that 31% of examined use cases had reached full production, twice the share in its 2024 study.

What controls should businesses implement for AI voice automation?

Businesses should implement a defensible control framework covering consent architecture, agent behavior, vendor obligations, and evidence preservation. At minimum that framework needs pre-deployment call classification, documented consent and opt-out records, human escalation paths, and ongoing testing of what the AI agent actually says on live calls.

A workable version includes:

1. Pre-deployment classification: map every channel to its initiation type and risk tier before launch.
2. Consent architecture: capture, timestamp, and store consent and revocation status per number.
3. Agent controls: lock down voice settings, calling hours, frequency, and agent identification scripts.
4. Vendor controls: define data sharing and contractual indemnification terms, which shift cost but never fully remove exposure.
5. Human escalation: route ambiguous consent situations or distressed callers to a live person.
6. Monitoring and evidence preservation: log every call, script version, and suppression-list check for later proof.

Agxntsix's AI Infrastructure work ties consent status, suppression-list checks, and call records directly into a business's CRM so that evidence exists before a regulator or plaintiff's attorney ever asks for it.

What does the ruling mean for consent and opt-out requirements?

Consent and opt-out requirements are untouched by the Ninth Circuit's ruling. A business running AI voice or text campaigns still needs prior express consent tied to each number, an honored internal opt-out list, and alignment with the National Do Not Call registry, regardless of how a court classifies one video-attachment edge case.

That holds even if a business tries to route its AI workflow through SMS instead of voice: the Ninth Circuit has already confirmed that a text message is still a TCPA 'call,' per the court's January 2026 opinion, so switching channels does not clear the bar. Other circuits disagree on that threshold question; Troutman's analysis of a parallel decision notes the Seventh Circuit found text messages are not 'calls' under the TCPA, creating a live circuit split. A business texting or calling from California-based numbers should assume Ninth Circuit treatment applies and build consent and suppression-list checks accordingly, confirming jurisdiction-specific exposure with counsel before scaling an outbound program.

How should businesses classify calls by initiation behavior?

Businesses should classify calls by what happens at the moment of connection, not by the technology behind them. A call that automatically begins speaking with a synthetic or prerecorded voice when the line connects sits in the highest-risk tier, while content requiring the recipient to take an action, such as pressing play on a video, sits in a lower-risk tier.

A yacht charter operator qualifying inbound quote requests with a live AI agent sits in the top row of that table the moment the agent starts talking; the same operator sending a text with an optional recorded walkthrough video sits closer to the bottom. Sorting every communication type this way, before launch, is the single clearest way to apply Howard's reasoning without over-reading it.

Balancing Growth and Compliance in Enterprise Voice AI

Balancing growth and compliance in enterprise voice AI means scaling call volume only as fast as consent and monitoring infrastructure can keep pace. EY found that organizations with real-time AI monitoring were 34% more likely to report revenue-growth improvements, showing that governance and growth move together rather than in competition.

EY's broader data backs that finding: organizations with real-time monitoring were also 65% more likely to report cost-savings improvements, while ISG's 2025 enterprise-AI study found that only about one in four AI initiatives met their expected revenue-impact goals even as production use climbed. The businesses that scale past that wall tend to be the ones treating consent architecture and call monitoring as part of the build, not an afterthought bolted on before a regulator asks. Agxntsix is a member of the Claude Partner Network, Anthropic's partner program for firms deploying Claude in production, and builds that governance layer directly into the voice AI and AI infrastructure work it delivers under its 60-day ROI positioning, so compliance monitoring ships with the calling program rather than trailing it.

Sources

• TCPA Tracker: January-February 2026
• Ninth Circuit rules Telephone Consumer Protection Act does not ...
• Seventh Circuit Rules Text Messages Are Not 'Telephone ...'
• Ninth Circuit Holds that Texts Are 'Calls' Under the TCPA But ...
• TCPA Tracker - Kelley Drye & Warren LLP
• Best Practices in Today's Fast-Changing TCPA Environment
• Ninth Circuit Decision on 'Texts-Versus-Calls'
• Voicemail - JD Supra