Agentic AI set to turbocharge cybercrime automation

Trend Micro has warned that a new wave of "agentic AI" systems will automate cybercrime operations at a speed and scale that existing defences may struggle to match.

The cybersecurity company has published research that describes how criminals are starting to use networks of specialised artificial intelligence agents. These agents sit under centralised orchestration systems. They run many stages of digital attacks with limited human direction.

The report focuses on what Trend Micro calls "VibeCrime". This is a model of cybercrime that relies on AI agents that act autonomously and in coordination with each other.

Trend Micro's Forward-Looking Threat Research team produced the study. The group tracks emerging criminal techniques and the structure of underground markets.

The researchers say agentic AI will change how criminals plan and execute campaigns. They say it will affect both traditional ransomware and phishing and newer forms of fraud.

Automated campaigns

The report states that attackers will chain together specialised AI agents under a central criminal orchestrator. Each agent will handle a task such as reconnaissance, message generation, data parsing, or payment processing. The orchestrator will assign tasks, monitor outcomes, and adjust behaviour based on results.

This approach mirrors trends in legitimate AI engineering. Many firms now use orchestration layers that manage multiple AI models and tools. Criminals are expected to adapt the same pattern inside their own "as-a-service" underground economy.

Trend Micro expects automated phishing, fraud, and exploitation of stolen data to run constantly. These operations will sit in the background and will not need frequent human input.

The company's researchers say this will increase attack volume. They also expect faster iteration of attack methods as AI systems test and refine different approaches.

The study includes two case studies. One describes how criminals might use agentic AI to parse large volumes of data from ransomware breaches. Another looks at how they could pair AI with licence plate recognition systems for targeted phishing schemes.

The authors say these examples are based on how criminals already use AI today. They also project how current techniques might evolve as agentic tools mature.

Shift in cybercrime model

The report predicts a shift from "cybercrime-as-a-service" to what it calls "cybercrime-as-a-servant". In this model, criminals rely on autonomous agents rather than human subcontractors for many tasks.

Ransomware groups today often rent infrastructure and tooling. They also rely on affiliates that run intrusion and extortion operations. Agentic AI could reduce the need for skilled affiliates. It could also lower the barrier for new entrants who lack technical expertise.

Trend Micro argues that criminal groups will treat AI agents as staff inside a virtual organisation. Each agent will follow goals that the orchestrator sets. The agents will also communicate with each other and adapt when defences block their actions.

Robert McArdle, Director, Forward-Looking Threat Research, Trend Micro, said agentic AI would change the speed and resilience of criminal operations.

"Agentic AI gives criminals a using ready-made arsenal that scales, adapts, and keeps working even when the humans disappear. The real risk is not a sudden AI-fuelled explosion of crime, but the slow, unstoppable automation of attacks that used to require skill, time, and effort. This shift is already underway," said McArdle.

The report suggests that this type of automation will affect the economics of cybercrime. Attacks that once took days or weeks of manual work could run continuously. Lower costs could make smaller or more marginal schemes viable.

Trend Micro also outlines what it calls the "Three Laws of Cybercrime Adoption". These principles aim to forecast how quickly criminals will adopt new tools. They cover short-term, medium-term, and long-term impacts of agentic AI on the threat landscape.

The researchers expect a near-term optimisation of current top-earning attacks such as ransomware and business email compromise. The medium term may see scaled-up exploitation of activities that are currently less profitable. The longer term may produce entirely new crime models that do not yet exist at scale.

Defensive response

The company says defenders will need to adjust architectures as criminals shift to agentic AI. It expects security platforms to add their own orchestration layers and autonomous defensive agents.

These defensive agents would monitor networks and user activity. They would also respond to anomalies and co-ordinate with each other without waiting for human analysts.

Trend Micro warns that current detection and response processes may not keep pace with new attack categories. It expects new forms of AI-enabled crime to appear faster than most organisations can detect and mitigate them.

McArdle said organisations face a strategic choice as criminals industrialise AI-based attacks.

"We will see an optimisation of today's leading attacks, the amplification of attacks that previously had poor ROI, and the emergence of brand new 'Black Swan' cybercrime business models," said McArdle.

He said this shift would require changes in corporate security planning and investment.

"For enterprises, this means reassessing security strategy now as well as investing in automation and AI-driven defence. Organisations also have to ensure resilience before criminals industrialise their own use of AI, or risk trying to catch up in an exponential arms race that will quickly separate those who were prepared and those were not," said McArdle.