Scaling Laws Caleb Withers on the Cybersecurity Frontier in the Age of AI
Dec 2, 2025
Caleb Withers, a research associate at the Center for a New American Security, dives into the impact of AI on cybersecurity. He discusses how frontier models shift the advantage to attackers, increasing both the scale and sophistication of cyber threats. Withers highlights persistent vulnerabilities in software and emphasizes AI's dual role in secure coding—both as a source of risk and a tool for detection. Additionally, he explores the urgent need for policy attention and industry responses to tackle the evolving landscape of AI-enhanced cyber warfare.
AI Snips
Chapters
Transcript
Episode notes
Frontier Models Are More Than Language
- Frontier AI refers to large foundation models that are multimodal and can effectively use computers beyond just language tasks.
- These broader capabilities make them more relevant to cyber operations than traditional 'language-only' framing suggests.
Cost Dynamics Could Shift Defender Edge
- AI still helps defenders, but frontier models change cost dynamics by keeping returns to scale for bigger or longer runs.
- That shift could make running defensive AI materially expensive and alter traditional defender advantages.
AI Amplifies Attack Scale And Quality
- Frontier AI amplifies both quantity and quality of attacks, making existing vulnerabilities easier to exploit at scale.
- Attackers willing to pay inference costs can probe more vectors than defenders can economically cover.