New Report Shows Intent, Not Appearance, Defines Risk: Nearly 97% of Instinct Sessions Presented as Adversarial Yet None Linked to an Attack
SAN MATEO, Calif.--(BUSINESS WIRE)--Arkose Labs, the leader in agent trust and orchestration, today announced The State of Agentic AI Trust Report, Q3 2026. It reveals personal agents’ rapid transition from reading the web to transacting on it. More than 65% of customer sites including banking, gaming, technology and travel organizations saw agents logging in, signing up and checking out on behalf of people. Meanwhile, controls haven’t caught up, with fewer than 3% of these agents proactively disclosing who they are.
“At a recent banking industry summit, the CISO of one of the world’s largest banks said that, for the first time in his career, a technology has arrived ahead of the controls," said Kevin Gosschalk, founder and CEO of Arkose Labs. "The State of Agentic AI Trust Report echoes this and makes it clear how difficult it has become for companies to distinguish legitimate, revenue-driving agents from malicious ones."
Drawing on production traffic from July 1 to September 30, the report’s key findings include:
- Muse changed the agent market in a matter of weeks. Sessions from Meta's Muse, a personal AI agent purpose-built for consumer use, grew more than 30x in September. This far outshadowed other agents, most of which held roughly flat.
- Hostile-looking is not the same as hostile. Instinct's design made 96.8% of its sessions look medium or high risk because it uses techniques commonly used by bad actors. Yet neither Instinct nor Muse had a single session linked to an attack.
- Agents look like bots, but they shouldn’t be treated like them. 77.4% of AI agent sessions were classified as legitimate, compared with just 7.9% of known bot sessions.
- “Allow all” or “block all” won’t work anymore. Agents use the same flows attackers target: 70.5% of registration sessions were fake-account attempts. Businesses can’t wave all automated traffic through or shut it all out; doing so turns away real customers and the revenue they bring.
- Most good agents don’t disclose themselves. 97.7% of legitimate AI agent sessions carried no verified disclosure.
- Agents are reaching the entire customer journey. AI agents appeared on login and registration flows in all industries measured. By the end of Q3, 65.2% of Arkose Labs customers saw them on their protected flows.
Part of what makes AI agents so difficult for companies to decipher is they often present themselves as something else. For example, Meta’s Muse appears on networks as a Chrome browser. Through a combination of behavior detection, biometrics, and insights from its global intelligence network, Arkose Labs is able to recognize three populations of agents: self-disclosing good agents, non-disclosing good agents, and malicious agents. Muse is a great example of a non-disclosing good agent.
“Every agent we track shows signs of automation somewhere. This was never the question,” said Mitchell Davies, lead researcher, Arkose Cyber Threat Intelligence Research (ACTIR) team. “The real question is, what did the agent come to do? What is its intent? Each company must decide how to handle them accordingly, taking their industry and risk posture into consideration. For example, a bank will have different controls in place than a retailer.”
The State of Agentic AI Trust Report, Q3 2026, is available for download today.
About Arkose Labs
Arkose Labs is the global leader in agent trust and orchestration, helping businesses grow revenue and stop fraud at scale. The unified Arkose Titan platform enables customers to identify visitors, starting with AI agents, and set policies governing what they can do and how they can safely transact. It also protects against fraud from any source, including account takeovers, fake account creation, and automated attacks. For more than 10 years, Arkose Labs has protected financial services, technology, gaming and travel companies, including Meta, Roblox, and Expedia. The company has been named to the Deloitte Technology Fast 500 for five consecutive years and backs Arkose Titan with the industry’s only $1 million warranty program. Arkose Labs is headquartered in San Mateo, California, with research teams worldwide.
Contacts
