Bryce Boland, Head of Security Solution Architecture, APJC with AWS. Handout
MANILA — Amazon Web Services on Tuesday touted the readiness of its systems to handle rising cybersecurity risks amid the emergence of AI agents using high-performance frontier models.
During a virtual briefing with tech journalists from Southeast Asia, Bryce Boland, Head of Security Solution Architecture, APJC with AWS, said the emergence of AI agents has fundamentally changed the landscape of cybersecurity.
Boland said AI agents have transformed how attacks can be implemented in scale, speed, and accessibility.
“The first is the difference in scale. In the past, vulnerabilities were found one at a time. They were sequentially found. Now we have high-performance AI models that can read entire code bases, and they can surface multiple or even large numbers of vulnerabilities simultaneously,” Boland said.
“The second is the difference in speed. High performance AI models enable attackers to find vulnerabilities very quickly,” he added.
“And third is the difference in accessibility. In the past, high risk vulnerabilities were typically discovered by very specific hacker groups or individuals that had deep specialized expertise, but AI has become a tool that anyone can easily access,” he said.
Boland said threat actors are taking advantage of this and using AI offensively today.
He also cited the arrival of Anthropic’s latest AI model, Claude Mythos, which reportedly could seek and exploit “zero-day” flaws in any IT operating system and web browser if asked. AWS said models like “Claude Mythos can now find software vulnerabilities and reason through complex attack paths at machine-speed, leading to an exponentially increasing backlog of vulnerabilities.”
AWS, the world’s biggest cloud computing service provider, has developed tools to help clients keep their data and systems secure amid the vastly altered threat landscape, Boland said.
“At AWS, we operate at a scale where we see threats long before most organizations will encounter them," Boland said.
AWS clients can rely on a platform called Continuum, previously called AWS Security Agent, which manages the full life cycle of a vulnerability at machine speed, he added.
“Continuum is a frontier agent that automatically reviews application security, finds vulnerabilities, and can even conduct context-aware penetration testing on demand. It supports the shift left narrative, finding security problems earlier in the development process,” Boland said.
Amazon said that Continuum addresses the complete lifecycle of code security at machine speed—from initial discovery through to actionable remediation.
The platform, which is now available in gated preview, offers threat modeling, design review, code review, and penetration testing.
Boland said AWS Continuum functions like a threat analyst by prioritizing findings depending on business impact. It also works like a Red Team Operator by verifying whether a vulnerability is genuinely exploitable and mapping out the attack path. And finally it also works as a security engineer by generating a tested fix with a rollback path.
But as models become more powerful, it becomes more important for developers of AI systems to ensure that proper guardrails are set up to ensure that they don’t threaten cybersecurity.
“AI systems themselves, when correctly engineered with the appropriate guardrails and controls and limitations on access, can be deployed in a very secure way, and that's what we've been observing,” Boland said.
He also said that right now, AI is a tool that is wielded by humans.
“I think behind every AI system there's intent, and intent is usually decided by humans. Today we see attackers already using AI systems to conduct various types of offensive operations. And fundamentally, that is human action, augmented and automated, using AI.”