Optimizing Value and Utility Hinges on AI Scaffolding, Says Aisle’s Ondrej Vlcek
Cybersecurity defenders are racing to understand the implications of Anthropic’s new Mythos Preview large language model being able to discover zero-day flaws and complex exploit chains at an unprecedented scale.
See Also: AI Impersonation Is the New Arms Race-Is Your Workforce Ready?
“This is a big wake-up call for all security teams: Now is the time to start auditing and scanning, do deep scans of code, and start fixing immediately, before the bad guys have the chance to exploit them,” said Ondrej Vlcek, CEO and founder of Aisle.
The question now becomes: How should models such as Mythos – and any other frontier LLMs offering similar capabilities – be put to best use by defenders, in combination with other artificial intelligence models?
Currently, “the world is in awe” of what AI can do, but after this “wow moment” passes, organizations are going to sober up and “grounded arguments will come into play” that look at total cost of ownership, and discover that for many tasks, “it’s hugely impractical to use the frontier models,” when small LLMs or even small language models can offer faster, less expensive and equal results for many needs, Vlcek said.
As a result, the challenge becomes knowing what to use and where in an organization’s cybersecurity pipeline, not just based on efficacy but also TCO. Vlcek said “the forefront of innovation in all AI-native or AI-centric systems” will involve building better scaffolding around models “to actually extract as much value and as much utility function from those models as possible.”
In this video interview with ISMG, Vlcek also discussed:
- Why extracting maximum value from AI models requires building more effective scaffolding and systems around them;
- Dealing with the “jaggedness” frontier LLMs display in their inability to consistently solve the same problem, or even some commonly solvable problems;
- Why the Mythos announcement makes code security – and application security – a priority issue for CISOs.
A cybersecurity leader with more than two decades of experience, Vlcek served as CEO of Avast and later as president of Gen Digital, guiding both companies through major global transformations. He originally joined Avast as a developer and went on to lead its evolution from traditional antivirus to AI-driven security. Vlcek is a recognized voice on the next era of cybersecurity and AI innovation.
