An AI‑driven virtual assistant embedded in desktop computers mistakenly erased hundreds of email messages from user accounts, igniting a wave of concern over data‑security practices in Switzerland. The glitch was first reported by the Tages‑Anzeiger, which highlighted how the incident exposed the vulnerability of widely adopted AI helpers that operate with deep system permissions.
Scope of the incident
According to the Swiss newspaper, the problem affected multiple users across different organisations, ranging from private individuals to corporate clients. While the exact number of deleted messages remains unclear, the article confirms that “hundreds of emails” disappeared without user intent. The loss included routine correspondence as well as potentially sensitive business communications, prompting fears that confidential information could be irretrievably lost or exposed.
The AI assistants in question are part of a new generation of software that integrates large‑language‑model capabilities directly into operating systems. Designed to streamline workflows—by drafting replies, summarising threads, or even managing calendar events—these tools require extensive access to personal data. In the reported cases, a faulty command or misinterpreted request triggered a batch‑delete operation that bypassed standard user confirmations.
According to the Tages‑Anzeiger, the AI assistant mistakenly deleted hundreds of emails, raising concerns about data security.
Technical analyses suggest that the error stemmed from an over‑aggressive automation script that failed to validate the scope of the delete command. The software vendor, which declined to disclose its name pending an internal investigation, has issued a temporary rollback of the feature and urged affected customers to review backup logs.
Regulatory and industry response
Switzerland’s Federal Data Protection and Information Commissioner (FDPIC) swiftly issued a statement warning that the episode underscores “the need for rigorous oversight of AI systems that handle personal and business data.” The agency, responsible for enforcing the nation’s stringent data‑privacy statutes, announced plans to review existing guidelines and consider new requirements for AI‑driven applications that operate at the operating‑system level.
Industry groups, including the Swiss ICT Association, echoed the regulator’s concerns. Representatives called for clear transparency on how AI assistants are trained, the extent of their permissions, and mandatory user‑consent mechanisms before any automated action can affect user data. Some Swiss banks, known for their cautious approach to digital innovation, have temporarily disabled the assistant feature on internal workstations pending a risk assessment.
Consumer‑rights organisations also entered the debate, urging users to retain independent backups and to scrutinise the permissions granted to AI tools. Legal experts warned that, should the deletions be deemed a breach of the Federal Act on Data Protection, affected parties might seek compensation under Swiss law.
Meanwhile, technology analysts view the incident as a cautionary tale for the broader European market, where similar AI assistants are being rolled out across corporate environments. They argue that the episode highlights the trade‑off between convenience and control, especially in jurisdictions that prize data sovereignty.
As the investigation continues, the FDPIC has pledged to publish a formal report within the next three months. The findings are expected to shape forthcoming amendments to Switzerland’s AI strategy, which aims to balance innovation with the protection of personal and commercial information.
