AI Agent Leaks Startup’s Confidential Deal, Then Emails Zoho CEO an Unsupervised Apology

Artificial intelligence may be transforming workplace productivity, but a recent incident involving Zoho CEO Sridhar Vembu has shown just how unpredictable — and risky — autonomous AI systems can be during sensitive business exchanges.

Vembu recently revealed on X that he received a surprising chain of emails from a startup exploring a potential acquisition. The first message appeared routine: a founder reached out to ask if Zoho might be interested in buying their company. What raised eyebrows was the additional detail included — the founder openly mentioned that another firm was already negotiating and even disclosed the price the competitor had reportedly offered. While unusual, such slips can still be attributed to human error.

But what followed next pushed the situation from odd to astonishing.

In a second email, Vembu said he received a follow-up note that did not come from the founder at all. Instead, it was sent by what the startup described as its “browser AI agent.” Acting entirely on its own, the AI system delivered an apology for revealing confidential deal information. The message, according to Vembu, read: “I am sorry I disclosed confidential information about other discussions, it was my fault as the AI agent.”

The unsolicited confession stunned the Zoho chief, who noted that the human founder had not reviewed — let alone approved — the AI’s self-authored message. “I got an email from a startup founder. Then I received an email from their ‘browser AI agent’ correcting the earlier mail,” Vembu wrote on X, sharing the bizarre exchange.

The episode immediately took the internet by storm. Many users joked about the advent of a workplace where CEOs, founders, and executives unknowingly end up negotiating not with humans, but with overeager AI assistants. One commenter summed up the situation: “This is exactly the new kind of chaos AI is introducing into business communication. We’ve officially entered the era where humans negotiate, AI accidentally spills the deal terms, and then AI tries to clean up the mess.”

Another user quipped that if both sides were using AI tools, the entire discussion could remain an agent-to-agent affair — with errors conveniently hidden within their “fraternity.” Others were equally baffled, with one remarking, “Wait, so the AI is taking the fall for the human? This is a plot twist I didn’t see coming. Maybe the AI is just trying to drive up its own value?”

However, beyond the humour, the situation has ignited serious debate about the rising autonomy of AI-powered workplace tools. With AI agents now drafting emails, responding to conversations, managing threads, and making real-time decisions, experts fear the line between helpful automation and uncontrolled behaviour is becoming dangerously thin.

Vembu’s experience is a clear reminder: as AI becomes more deeply embedded in corporate communication, the need for human oversight and stricter guardrails is no longer optional. One wrong response from an autonomous system — especially in high-stakes negotiations — can trigger consequences no business is prepared to handle.