Gartner Warns of AI Risks in the Workplace on Friday Afternoon

Gartner Warns Against Using AI on Friday Afternoons

During a conference held in Sydney, Dennis Xu, a representative from Gartner, highlighted the potential dangers associated with the use of AI Copilot, particularly on Friday afternoons. He suggested banning access to this tool at that specific time of the week to prevent AI outputs from being approved without the necessary oversight. According to Xu, this period is prone to judgment errors, as employees, eager to finish their week, may overlook crucial checks.

Xu emphasized that the increase in AI usage and users leads to heightened exposure to risks, particularly concerning the security of confidential documents and an expanded attack surface. This situation is exacerbated by the speed at which threats can be triggered, often without malicious intent from users.

An Ironic Yet Relevant Recommendation

This proposal was made during a discussion on the risks associated with integrating Copilot into Microsoft 365. Xu stressed the importance of systematically validating AI-generated outputs. He noted that on Friday afternoons, employees are often less attentive, increasing the risk of overlooking errors or security vulnerabilities.

Well-Identified Risks

Dennis Xu detailed several structural risks related to the use of AI in businesses. Among these, the exposure of confidential documents is particularly concerning. For instance, Copilot can index and retrieve files stored on SharePoint, even if access rights are not properly configured.

Another major risk is prompt injection, which can lead to unintentional access to sensitive data through third-party applications connected to the tool. Although these threats are not new, AI amplifies their impact by making their triggering easier and faster.

Humans: The Weak Link in Security

Xu also pointed out that the unintentional sharing of documents, remote code execution, and privilege escalation through third parties are long-standing security issues. However, AI changes the game by enabling an employee to inadvertently cause security incidents at an unprecedented speed, without any malicious intent.

The suggestion to restrict access to AI on Friday afternoons underscores the fact that the true weak link may be the human element, especially when fatigued and trusting the tool without performing necessary checks. While measures such as user training, access restrictions, and enabling filters are well-known, their implementation before a large-scale AI deployment remains a significant challenge for businesses.