This week brought serious warnings about AI agent safety from major technology companies and research groups. A new report from Salt Security found that most businesses are not ready for the security risks that come with AI agents. These smart computer programs can make decisions on their own, but they also create new ways for hackers to attack companies.

The biggest problem is that API security - the way different computer programs talk to each other - is not strong enough. Over half of companies using AI agents plan to let them talk directly with customers, but only 32% check for security problems every day. This creates a trust gap where people worry about sharing personal information with AI systems.

Another study by Infosys found that 95% of business leaders have seen problems with their AI systems in the past two years. Almost 40% said these problems caused severe damage to their companies. Even worse, 86% of executives believe AI agents will create new risks, but only 2% of companies have proper safety controls in place.

Experts warn that companies are moving too fast to use AI agents without thinking about safety first. In Asia, organizations are rushing to deploy AI systems to stay competitive, but security is falling behind. The report suggests that security must be built in from the start, not added later as an afterthought.