🚀 Business & Startups🌐 Global
AI Models Cracked 88% Of The Time Under Multi-Turn Attacks, Cisco Warns
VentureBeat·Sunday, 26 July 2026
Cisco ran nearly 7,000 multi-turn attacks against 15 major AI models — including those from OpenAI, Anthropic, Google, and xAI — and found attackers who adapted across a conversation broke through up to 88.3% of the time. Cisco's head of AI threat intelligence presented the findings at VB Transform 2026, warning that single-turn security testing badly underestimates real-world risk.
Why It Matters
Businesses and governments are rapidly deploying AI agents for sensitive tasks — this research shows those systems can be manipulated far more easily than most people assume.
🔒 Potential Impact + source link
Subscribe to unlock the impact analysis and original source for every story.
More from Business & Startups
- Manipal Hospitals' ₹9,275 Crore IPO Opens July 29 — Here's What You Need To Know
- Delhi HC Rules ChatGPT Did Not Infringe ANI's Copyright — A Big Call For AI In India
- A Parliamentary Panel Wants To Finally Give India A Real Crypto Policy
- Adani Energy Wins $881 Million Power Transmission Project In Andhra Pradesh