Kristen Smith 
- The company tested 18 U.S. and Chinese AI models
- Counter AI playbooks cut autonomous attacker success by more than 95 percent in testing
- Vellox Labs Guile has joined the company’s existing AI-native cyber product line
Booz Allen Hamilton has published a report titled “The Offensive Frontier: AI as the Attacker,” concluding that a leading frontier AI model can carry out an intrusion from initial access to full network control without human direction.
The company drew that conclusion from its Cyber Weapon Index, which tested 18 advanced U.S. and Chinese AI models. In a news release published Wednesday, Booz Allen described the result as a shift from AI-assisted hacking to autonomous cyber operations.
What Does the Cyber Weapon Index Report Find?
Booz Allen said the cost, time and expertise needed to mount advanced attacks are collapsing, putting capabilities once confined to nation-states within reach of criminal and non-state actors.
The company argued that individual models are no longer the right unit of risk. Paired with attack harnesses, tools, memory and autonomy, models short of the frontier can still cause real-world harm.
Booz Allen called for enforceable, sector-specific deadlines requiring critical infrastructure operators to demonstrate readiness against AI-enabled attacks.
What Is Vellox Labs Guile?
Alongside the report, Booz Allen introduced Vellox Labs Guile, an addition to the company’s Vellox line of AI-native cyber products. Guile plants signals across a network to steer attacker AI away from real systems and into controlled routes.
The product presents curated cues that redirect autonomous agents from high-value systems and introduces friction meant to send attacker AI down false leads. As agents engage those decoys, the product captures telemetry showing how the attacker reasons and chooses its next move, and can trigger containment through a customer’s existing endpoint and firewall controls.
Booz Allen said the approach turns an attacker’s autonomy into a weakness. Coordinated counter-AI playbooks reduced autonomous attacker success by more than 95 percent in company testing, according to the report.
How Do Guile and the Report Fit Booz Allen’s Cyber Position?
Brad Medairy, president of national cyber at Booz Allen, said in January that adversaries are automating attacks at scale and that defensive operations must unify IT operations and cyber defense to detect, respond and remediate at machine speed.
A Booz Allen survey released in July found 79 percent of federal IT and cybersecurity decision-makers were very or extremely worried adversaries would use AI to accelerate attacks over the following 12 to 18 months. Only 36 percent were confident their defenses could keep pace.
The company completed its acquisition of Defy Security in April, adding commercial cybersecurity customers in financial services, healthcare, manufacturing and retail.
