A Quarter of UK and US Firms Suffer Data Poisoning Attacks

British and American cybersecurity leaders are raising alarms about the rapidly expanding attack surface created by artificial intelligence, highlighting that risks are evolving as quickly as the technology itself. A significant concern is the prevalence of shadow AI, where employees use unsanctioned generative tools, introducing severe risks of data leakage and compliance violations. This unauthorized usage often bypasses security protocols, exposing organizations to vulnerabilities found in public models. Equally troubling is the rise of data poisoning attacks, which were previously considered largely theoretical but are now affecting over a quarter of enterprises. These attacks interfere with model training data to alter AI behavior, potentially sabotaging critical infrastructure or causing security systems to fail. This underscores a critical vulnerability in how organizations ingest and process information, threatening the fundamental integrity of the services they rely on. This article is highly relevant to open_data because it demonstrates how the integrity of publicly available datasets is under direct threat. If training data can be easily poisoned or if open tools are used without governance, the reliability of open datasets used for AI development becomes questionable. It emphasizes the urgent need for robust governance and validation processes to protect open data ecosystems from manipulation and misuse.

Source: infosecurity-magazine.com
Published on 2025-09-18

A Quarter of UK and US Firms Suffer Data Poisoning Attacks

Related news