When Forrester CEO George Colony took the stage at the Forrester Security and Risk Forum 2026 in Washington, D.C., on Nov. 9, he declared that the “security singularity” is here and challenged security leaders to rethink how they protect organizations as AI agents gain increasing levels of autonomy.
The security singularity, according to Colony, describes a point at which emerging technologies such as AI fundamentally change the assumptions underpinning traditional cybersecurity.
“The way we’re trained will change,” he said. “The vectors of attack and how we defend will change. How we write software will change,” he said.
Change was the central theme throughout the opening remarks at the conference. From an AI-driven reimagining of chess strategy to the challenge of securing machine intent, speakers outlined what organizations must rethink as AI agents take on greater autonomy.
AI changing how humans play the game
In his opening remarks, Forrester VP and research director Merritt Maxim spoke about the dawn of computer chess, drawing an analogy between how AI transformed chess strategy and how it is now reshaping cybersecurity. After years of development, advanced neural networks and chess engines such as AlphaZero became so capable that they have fundamentally changed how humans approach the game.
Traditional chess strategy emphasized protecting the king. AI-powered chess engines challenged that assumption, demonstrating that more aggressive approaches could produce better results, using the king in offensive tactics.
Chess players have since adopted similar playing styles, said Maxim. “The result was that long-term king safety was not a feature that they employed and it actually changed the way humans approached chess. So, I think it’s a future analogy about how computers can coexist with humans and ultimately drive better performance.”
Maxim suggested AI will not simply help cybersecurity professionals perform existing tasks faster. It will change how humans approach cybersecurity itself. Human judgment will remain essential, particularly when dealing with the nuance and complexity of real-world business environments. But security teams, Maxim said, must also learn to work alongside increasingly capable AI systems and rethink established practices.
Becoming an agentic company requires securing intent
One of the day’s most significant takeaways was the need for organizations to understand and secure AI intent.
In his keynote, Forrester VP and principal analyst Jeff Pollard highlighted the distinction between task adherence and goal adherence as a particularly important security challenge: an agent might successfully complete its assigned task while pursuing the broader business goal in ways its developers never anticipated.
He explained that understanding intent requires grasping three interconnected elements: the objective assigned to an agent, the constraints and boundaries placed around it, and the actions it takes over time.
He provided an example of an accounts receivable agent instructed to accelerate customer payments. Initially, it sends routine payment reminders. But if customers remain unresponsive, the agent might search social media for additional information, personalize communications and eventually publish public accusations about customers that fail to pay.
“No one thought that was going to happen, but it’s a problem,” said Pollard. “We have this issue where we have to balance the security question of task adherence — did it do what we wanted it to do? — with the governance question of, did it accomplish the goal that we wanted it to accomplish? This is part of the underlying question of securing intent.”
Striking the balance between security and AI’s potential
Pollard also warned against overcorrecting for the risks associated with autonomous AI.
Security teams naturally focus on accidental harm and purposeful malicious behavior. But doing so risks overlooking what he described as engineered and emergent helpfulness: the capabilities that make AI valuable.
AI agents can enable employees to build things, solve problems and accomplish work that previously would have been impossible or prohibitively expensive. He warned that overly restrictive security controls could stifle those benefits and prevent organizations from realizing the potential of agentic AI.
A core message from the opening day was that agentic AI is advancing too quickly for organizations to rely on traditional security models alone. Security leaders must develop visibility into agent behavior, understand the objectives and constraints governing autonomous systems, and continuously assess the risks their actions create.
As Pollard emphasized, the goal for any organization attempting to fully embrace autonomous AI is to understand intent and ensure agents act with trusted autonomy.
“That is agentic intent,” he said. “Put simply, it’s what the agent was asked to do, what the agent did, what the agent didn’t do, and what happened as a result. That’s the sum total of intent. Your enterprise cannot become agentic and, therefore, it cannot become autonomous without understanding intent.”
Richard Livingston is an editor for TechTarget Cybersecurity, covering news, trends and analysis. Livingston’s professional background includes editorial positions in the national defense industry covering the U.S. Army Medical Department, as well as offensive and defensive cyber strategy for military and government audiences.
