AI & the Death of Accuracy: What It Means for Zero-Trust

AI “model collapse” could degrade large language models as they train on increasingly AI-generated data, potentially introducing inaccuracies and eroding PII protections, according to the article. Gartner on 21 January predicted that by 2028, 50% of organisations will adopt a zero-trust data governance posture due to unverified AI-generated data, a concept the piece labels as model collapse.

The piece notes that LLMs are trained on data scraped from the Internet and other sources, including AI-generated content, which can degrade model quality and lead to confident-sounding but flawed outputs in areas such as code reviews and patch recommendations. It quotes Gartner saying that 84% of CIOs and Technology Executives expect to increase GenAI funding in 2026, while highlighting regulatory pushes to verify AI-generated data and enforcing a zero-trust posture for data governance.

The discussion also features insights from AppOmni and Noma Security on data provenance and continuous model behaviour evaluation, emphasising the need to treat training data as a governed asset. 27 January 2026.