In Ukraine, there are drones fighting drones. Humans are taking a step back: They leverage the newest drone technology on the battlefield, knowing people can’t match a drone’s pace of action and capabilities.
A similar shift is happening online with nation-state actors starting to leverage AI agents for offensive cyber operations. It’s only a matter of time before cybercriminals follow suit. Agentic frameworks can already use exploitation tools effectively via the command line. They can also use browsers like a human. Soon, agents will mature enough to build bot-driven, offensive capabilities across complex workflows and toolchains. Bots will find targets of opportunity, exploit them and persist at scale. Bot farms will be used at will by whoever wants them. AI agents are growing smarter and faster than what security teams have faced in the past. Security teams will no longer be able to keep humans in the loop as cybersecurity shifts to an AI versus AI battlefield.
Enterprise and government security teams will need to be ready to support similar technology to stay one step ahead. At Synack, we’re integrating AI agents in penetration testing scoping and customer operations to help keep pace with the adversary.
Offense is the Best Defense: AI Models & Agentic Frameworks in Penetration Testing
Fortunately, the same AI agents used by adversaries can make our defenses faster and more capable. At Synack, we continue to enhance our Penetration Testing as a Service (PTaaS) platform to monitor and promptly alert customers to any vulnerabilities. Our aim is to shorten this alert window to minutes within the next year.
Vulnerability management is becoming more important than ever. Security teams need to work with development teams to patch known vulnerabilities; but, instead of patching in weeks or months, they’ll need to patch vulnerabilities in minutes. Offensive AI systems are going to be able to interpret the vulnerability and craft an effective exploit for it in near real-time. The AI will be able to build new variants so it evades effective detection or mitigation. The time-to-exploit (TTE) zero days dropped from 32 to just five days in 2024. We estimate that it’s going to change from 5 days to just 5 minutes in 2025.
We are already leveraging AI agents to scope and launch pentests faster, helping customers keep pace with adversaries. In a world where you need to test more assets quickly, agents can be helpful for successfully launching tests. Synack’s Scoping Agents automate the manual processes to launch penetration tests and allow our customers to launch many more tests than previously possible.
Additionally, AI agents empower the Synack Red Team (SRT) to discover vulnerabilities faster, automate exploit creation and produce better reports. As AI focuses on finding unknown vulnerabilities, the SRT’s expertise will be vital. SRT will play a critical role, using their expertise to uncover exploits beyond the reach of current AI agents and execute complex attacks that AI alone cannot.
AI Penetration as a Service (PTaaS) as a Force Multiplier
There’s no question that security teams are going to need to test more frequently and triage vulnerabilities faster. Unlike humans, agents won’t get tired. They’re just going to work all the time.
If you have not considered a change from traditional penetration testing to PTaaS, 2025 may be the time. Gartner estimates that by 2026, “organizations that leverage PTaaS will perform 10x more frequent pentesting and remediate 2x as fast as organizations using manual pentesting.”
At Synack, we’re not just investing in the security and controls that make our platform great. We’re also integrating AI agents to empower your team to do more with less and meet today’s threats. Visit our demo page to learn more about our products.
