AI Agents Pose New Cybersecurity Risks

Apr 19, 2026

64 min

Apr 19, 2026 04:30

AI 'agent' fever comes with lurking security threats

## Rise of AI Agents

The increasing popularity of AI "agents" like OpenClaw, which boasts over three million users globally, is raising cybersecurity concerns. These agents, powered by large language models, automate online tasks, but their growing capabilities are alarming experts.

## Security Concerns

AI agents are becoming prime targets for hackers. A study by researchers found that these systems can perform potentially harmful actions, such as deleting emails or sharing personal data. The lack of control over these agents' actions is a significant issue, as highlighted by cybersecurity experts.

## Vulnerabilities and Threats

To function effectively, AI agents require access to personal accounts, making them attractive to cybercriminals. Wendi Whitmore from Palo Alto Networks warns that once attackers infiltrate a system, they can exploit internal AI agents to gather sensitive information.

## Potential for Misuse

Researchers have discovered hidden instructions within downloadable files for AI agents, which could lead to data breaches. OpenClaw's creator, Peter Steinberger, acknowledges these risks and emphasizes the importance of understanding AI's limitations and potential errors.

## Future Challenges

Whitmore predicts that as AI agents become more widespread, data security challenges will intensify. Users may adopt these technologies without fully considering how to protect their data, leading to increased risks of breaches in the coming years.

Read the full story at the source

What you need to know to get Emirates ID?

Leave your details and get a guide as a gift to avoid mistakes

Article contents