'It was my fault': AI agent leaks biz secrets to Zoho’s Vembu, then apologises

New Delhi: Zoho Corp founder Sridhar Vembu on Friday shared an amusing account of how a startup’s AI agent accidentally spilt sensitive information and later apologised — underscoring the growing, and at times chaotic, role of autonomous AI tools in high-stake corporate communications.

According to Vembu, who took to X to recount the episode, it all began when he received a cold e-mail from a startup founder pitching a potential acquisition.

However, the e-mail contained information typically guarded with utmost secrecy during M&A (Mergers and Acquisitions) discussions — the name of a competing bidder and the specific price they had offered.

Revealing a competing offer is considered a high-risk negotiation tactic, but disclosing specific details often breaches confidentiality agreements.

Things took a turn when Vembu received another e-mail — not from the startup founder, but from their “browser AI agent”.

The automated system reportedly sent a correction, stating: "I am sorry I disclosed confidential information about other discussions, it was my fault as the AI agent."

Vembu’s post quickly went viral, sparking debate among netizens about the dangers of Agentic AI — systems designed to execute tasks autonomously.

One user quipped, “We’ve officially entered the ‘please ignore my AI, it’s still learning NDAs’ era.”

“This is exactly the new kind of chaos AI is introducing into business communication. We’ve now entered a phase where humans negotiate, AI accidentally spills deal terms, and then AI tries to clean up the mess. It’s funny, but also a reminder — AI can assist in communication, but it can also miscommunicate, overshare, or breach confidentiality if not properly controlled,” another user commented.

The incident has raised serious questions about the lack of guardrails in current AI implementations. While firms are rushing to deploy AI agents for sales and scheduling, these models often struggle with context, nuance, and the strict confidentiality required in corporate deal-making.

“One slip, and suddenly the ‘helper bot’ is leaking acquisition terms. We’re entering a phase where AI won’t just draft emails - it’ll accidentally start negotiations,” one user observed.

Beyond the humour, industry watchers see the episode as a warning sign of potential corporate embarrassments to come. “This is going to become a very common failure mode,” one commenter cautioned.

“People will let agents handle ‘important emails’ and realise only later that they’ve leaked something that should never have been mentioned.”

However, not everyone was convinced that a rogue algorithm was solely to blame.

Sceptics suggested the “AI apology” might be a modern face-saving tactic, or a calculated bluff.

“There was a human behind that AI agent, and they just want you to think it was a mistake when it wasn’t,” one user remarked.