When AI fails: five cases that raised alarms, from a US$25 million fraud to autonomous agents
According to IBM and the Ponemon Institute's Cost of a Data Breach Report 2026, one in four malicious breaches analyzed was enabled by artificial intelligence, at an average cost of US$6 million versus US$4.99 million for the global average. The study examined incidents at 602 organizations between March 2025 and February 2026. The article reviews five cases: autonomous agents that tried to breach more than 1,200 websites, AI-assisted attacks detected in Taiwan, a US$25 million fraud through a fake video call, an adverse ruling against Air Canada over its chatbot, and Microsoft's Tay chatbot case.
In 45 seconds
- IBM and Ponemon: one in four malicious breaches analyzed was enabled by AI, at an average cost of US$6 million.
- An attacker used autonomous AI agents to try to breach more than 1,200 sites and infiltrated pages of over 100 companies, including Deere, SharkNinja and Cummins.
- Taiwan detected in July 2026 attacks on public agencies that combined human actions with AI agents.
Read the full story at La República
IASummary generated by artificial intelligence from a story by La República. It may contain errors: the original source prevails.
Comments
Be respectful. Comments reported by several people are hidden until a moderator reviews them.
You do not need an account to read, only to comment. Sign in with a code sent to your email or with your Google account. We only show the name you choose, never your email.