AI Agent Ruling to Boost Identity Verification Demand

AI Agent Ruling to Boost Identity Verification Demand

By Ian Tang and Kate Lardner
Capstone TMT Analysts
November 19, 2026

Capstone believes identity verification providers will benefit as online platforms allowing agentic artificial intelligence (AI) increasingly need to distinguish legitimate agents from fraudulent bots, reinforced by a US Court of Appeals for the Ninth Circuit ruling that treats AI agents as extensions of human users. Platforms will need to verify users linked to agents. Those that block AI agents (e.g., banks and some e-commerce firms) will rely on bot management vendors.

  • In Amazon v. Perplexity, the Ninth Circuit held that users, not their AI agents, “access” websites under the Computer Fraud and Abuse Act (CFAA), in an August 4th ruling. The court’s reasoning and recent regulatory signals treat AI agents as “tools” and point responsibility toward human users. Senator Mark Warner’s (D-VA) discussion draft of a Senate bill released in June requires large online platforms to allow agents when tied to verified human users.
  • Given these developments, we believe online platforms that admit AI agents will need to verify human identities behind them to separate authorized agents from fraudulent ones, driving demand for third-party providers of biometric verification, passkeys, digital IDs, and agent identity and access management.
  • We expect online platforms that block AI agents, such as some banking and e-commerce platforms, to rely on bot management and fraud prevention vendors.

Background

The emergence of agentic AI, which allows autonomous AI agents such as Perplexity AI’s Comet and Meta Platforms Inc.’s (META) Muse to browse the web and complete tasks on a person’s behalf, has pushed platforms to strengthen technical restrictions and, in some cases, legally challenge the conduct. Amazon.com Inc. (AMZN) is suing Perplexity AI, alleging Comet, the autonomous shopping tool, accessed its platform without authorization. In September, Amazon blocked Muse from its website.

However, early legal signals have started to treat agents as legitimate extensions of their users. On August 4th, the Ninth Circuit vacated Amazon’s preliminary injunction against Perplexity, ruling that it is the user, rather than the AI agent, that accesses the website. Separately, Senator Warner released a discussion draft of the Artificial Intelligence Access, Gatekeeper Exchange, and Nondiscriminatory Transfer Act of 2026 (AI AGENT Act, S. 5051) in June, which would require large platforms to allow AI agents on their websites as long as they are tied to a verified human user.

Capstone believes these developments, as well as the growing difficulty of distinguishing legitimate agents from fraudulent automation, are increasing online platform demand for verifying the individual behind an agent, alongside traditional bot detection and blocking, building on the regulatory pressures we have previously described. In this note, we discuss what has changed in this space and which types of identity verification providers are best positioned.

Amazon v. Perplexity AI

On August 4, 2026, the Ninth Circuit overturned a preliminary injunction Amazon had previously won in Amazon v. Perplexity, ruling that as human users direct the activity and communications, it is the user, not Perplexity, that accesses Amazon’s systems. The court found Amazon was unlikely to succeed on its CFAA and California state law claims, reasoning that AI agents are merely “tools,” and human users operating them are the true actors. The decision vacated a lower court’s March 2026 injunction that had halted Perplexity’s agents from accessing Amazon. The Ninth Circuit denied rehearing on September 10th. The case now continues in the US District Court for the Northern District of California, where Amazon filed an amended complaint on September 21st.

The Ninth Circuit ruling is narrow as it applies only to “access” definitions under the CFAA and Comet’s specific technical nature. It leaves open the possibility of other legal claims, such as breach of contract, terms of service, or tort claims. The ruling limits the claims plaintiffs can bring but does not provide broad legal immunity for agentic AI. We believe platforms are likely to continue bringing contract and other non-CFAA claims against agentic AI activity, but such claims will likely focus on individual users’ conduct rather than unauthorized access to computers or digital infrastructure.

Even when AI agents are treated as human-driven tools, online platforms can still ban their use via terms of service and technical blocks. Those that choose to admit them, however, especially in the commerce space, will be increasingly incentivized to verify the human identity behind agent activity. Verification will allow them to distinguish legitimate AI agents acting on behalf of real customers from fraudulent automation.

Warner AI Bill

On June 29, 2026, Senator Warner released a discussion draft of the AI AGENT Act of 2026. The legislation proposes giving individuals the legal right to use autonomous AI agents on major website platforms, limiting online companies’ ability to block AI assistants while imposing guardrails such as pre-deployment testing and an incident reporting system. To qualify, agents and the organizations deploying them would have to link each agent to a verified operator.

The draft remains in the earliest stages of the legislative process with slim chances of passage this session. Nonetheless, it signals that policymakers are moving from restricting AI agents to accepting that they are here to stay. Liability for unauthorized or divergent agent conduct is an open question in the draft bill. However, the Ninth Circuit decision suggests that, at least under the computer-fraud law, the legal focus for agentic AI accidents will largely fall on human users directing agents.

FTC Dynamics

The Federal Trade Commission (FTC) has signaled a similar view on liability. On September 25, 2026, FTC Chairman Andrew Ferguson said at a Reuters event that he would resist treating AI agents as autonomous beings with “wills and desires of their own,” treating them instead as tools. He suggested that those who instruct agents, including developers, could be liable for harms. The remarks are not formal guidance or rulemaking.

Investment Targets and Platform Incentives

Traditionally, websites and online platforms block unauthorized traffic by blocking IP addresses, flagging bot-like activities or requiring users to use CAPTCHAs. These methods focus on “what” is visiting a website, relying on technical clues. Because many AI agents often operate through human users’ internet browsers and IP addresses, they are difficult to detect, making traditional blocking less effective. We believe websites will increasingly add identity checks, targeting “who” is behind the screen, alongside traditional blocking.

We expect platforms and merchants that admit agents to increasingly tie agent activity to a verified human user to separate legitimate agents from fraudulent ones. These platforms, as well as AI developers, will increasingly employ third-party providers of:

  • Biometric verification (e.g., liveness detection and facial recognition)
  • Passkeys
  • Digital IDs
  • Agent identity and access management

However, platforms such as Amazon have multiple potential motivations for blocking agents:

  • Data access: Companies like Amazon do not want AI developers such as OpenAI or Anthropic to have access to their pricing data, which they consider competitive assets.
  • Advertising: Companies with advertising businesses like Meta, LinkedIn, and Amazon depend on humans viewing their sponsored content, while AI agents ignore advertisements.
  • Dark patterns: Some platforms rely on dark patterns such as difficult-to-cancel subscriptions, hidden add-ons, or false scarcity (e.g., “Only 2 left in stock!”). Those practices rely on human behavior, while AI agents would ignore them.
  • Bank fraud: For banks, the clearest incentive is fraud. When consumers give AI agents their bank information, the law says they have no redress to the bank for fraud. Still, banks often feel pressured to make consumers whole to avoid reputational risk. For example, on October 13th, Jamie Dimon, CEO of JPMorgan Chase & Co. (JPM), announced that the company would block external AI agents from performing automated transactions.
  • Customer relationships: Online platforms selling directly to consumers may block AI developers from interfering with that relationship. For example, on November 11th, United Airlines (UAL) announced that it would take steps to restrict AI agents from booking flights on its digital platform, with CEO Scott Kirby saying he does not want to create “another middleman.” By contrast, Delta Airlines (DAL), American Airlines (AAL), and JetBlue (JBLU) allow customers to book flights through agentic AI.

Blocking will create a separate demand for vendors of:

  • Bot management and agent traffic management
  • Fraud prevention

Businesses providing identity verification services will also see increased demand due to intensifying children’s online safety and data privacy concerns. We expect both established identity verification vendors and emerging entrants to be well positioned to capture share in this quick-moving market.

What’s Next

The judge in Amazon v. Perplexity is scheduled to hear Perplexity’s motion to dismiss Amazon’s amended complaint on November 20th. Discovery in the district court case is expected to close on February 12, 2027, with a jury trial scheduled for September 27, 2027.

Following the Ninth Circuit decision, Amazon said it is weighing further legal options, including Supreme Court review.

Read more from Capstone’s TMT Team:

Nvidia’s Political Reckoning
Regulatory Tailwinds Build a Case for Identity Verification Vendors
Why Cybersecurity Advisory Firms Will Stay Resilient Despite DOD Mandate Suspension

Have a question?

We want to hear from you. Let us know your question and a research analyst will get back to you promptly. We love to discuss our research.

Connect

Our Latest Insights

Nvidia’s Political Reckoning

Nvidia’s Political Reckoning

It took less than a week for Jensen Huang to go from one of President Trump’s closest allies to a political liability. Allegations that the now-former Nvidia CEO was targeted in a suspected Chinese espionage plot culminated in his resignation on Thursday. We believe...

New Power in Washington Puts Data Centers on the Hot Seat

New Power in Washington Puts Data Centers on the Hot Seat

The 2026 midterm elections were a seismic event, and we believe the incoming Democratic committee chairs will make conducting investigations and holding public hearings regarding data center policy in the next Congress a high priority. However, these investigations...