David Aviv on adversarial AI
From David Aviv’s interview with the Cyber Risk Alliance at RSAC 2025 · · Radware
“We're seeing agentic AI bots that, when given our JavaScript anti-bot challenges, forward them to large language models which quickly solve them — allowing attackers to bypass protections and scrape entire catalogs.”
On , David Aviv, CTO at Radware, spoke about adversarial AI during David Aviv’s interview with the Cyber Risk Alliance at RSAC 2025 on Radware.
About David Aviv
At RSAC 2025, Aviv stated that the democratization of AI is "changing dramatically not only the security posture but our lifestyle" and argued that human abilities will not keep up with AI abilities. He described seeing "agentic AI bots" that forward JavaScript anti-bot challenges to large language models, which quickly solve them and allow attackers to bypass protections. Aviv said Radware uses AI to build deeper security controls and to protect AI infrastructure, noting that running AI applications creates a new AI supply chain and new threat surfaces. In earlier appearances, Aviv discussed the shift to edge-centric architectures driven by 5G, stating that security must be pushed to the edge and that the 5G core, built on HTTP/web-based protocols, opens networks to web-based attacks. He noted that account takeover has gained momentum via spear-phishing, allowing attackers to assume identities and install stealth agents in cloud environments. Aviv also described Radware's approach to self-learning security algorithms, stating that in one week their devices produced about 130,000 attack events and roughly 99.9% of mitigations were performed automatically.