Tech »  Topic »  Securing AI Models Against Unpredictability and Exploitation

Securing AI Models Against Unpredictability and Exploitation


Cisco's Jeetu Patel on Open-Source Security and Building Safer AI Systems Michael Novinson (MichaelNovinson) • April 29, 2025

Artificial intelligence is transforming cybersecurity from the ground up. Security teams battle skills shortages, alert fatigue and bloated technology stacks. Jeetu Patel, Executive Vice President and Chief Product Officer at Cisco, said AI can augment capacity and simplify defenses when applied thoughtfully.

See Also: Dealing with Data Explosion Challenges and Pain Points

Patel warned that traditional architectures can't handle the inherent uncertainty of models, especially since many remain "jail-breakable," leaving systems vulnerable to exploitation.

"When DeepSeek came out, in the first 48 hours, we were able to go and have 100% attack success rate against DeepSeek with the top 50 categories identified in harm bench, which was a benchmark independent for AI models," he said. "You want to make sure that these models get validated and then there is runtime enforcement ...


Copyright of this story solely belongs to bankinfosecurity . To see the full text click HERE