{"slug": "microsoft-research-llms-corrupt-your-files-during-delegated-work", "title": "Microsoft Research: LLMs Corrupt your files during delegated work", "summary": "A new study from Microsoft Research, DELEGATE-52, found that 19 large language models, including frontier systems like Gemini 3.1 Pro and GPT 5.4, corrupt an average of 25% of document content during long delegated workflows across 52 professional domains. The research shows that errors compound over time, with degradation worsening due to document size, interaction length, and distractor files, while agentic tool use failed to improve performance. The findings indicate that current LLMs are unreliable delegates that silently introduce severe errors into documents, undermining trust in AI-assisted knowledge work.", "body_md": "# LLMs Corrupt Your Documents When You Delegate\n\nLarge Language Models (LLMs) are poised to disrupt knowledge work, with the emergence of delegated work as a new interaction paradigm (e.g., vibe coding). Delegation requires trust – the expectation that the LLM will faithfully execute the task without introducing errors into documents. We introduce DELEGATE-52 to study the readiness of AI systems in delegated workflows. DELEGATE-52 simulates long delegated workflows that require in-depth document editing across 52 professional domains, such as coding, crystallography, and music notation. Our large-scale experiment with 19 LLMs reveals that current models degrade documents during delegation: even frontier models (Gemini 3.1 Pro, Claude 4.6 Opus, GPT 5.4) corrupt an average of 25% of document content by the end of long workflows, with other models failing more severely. Additional experiments reveal that agentic tool use does not improve performance on DELEGATE-52, and that degradation severity is exacerbated by document size, length of interaction, or presence of distractor files. Our analysis shows that current LLMs are unreliable delegates: they introduce sparse but severe errors that silently corrupt documents, compounding over long interaction.", "url": "https://wpnews.pro/news/microsoft-research-llms-corrupt-your-files-during-delegated-work", "canonical_source": "https://www.microsoft.com/en-us/research/publication/llms-corrupt-your-documents-when-you-delegate/", "published_at": "2026-05-26 22:21:10+00:00", "updated_at": "2026-05-26 22:37:13.620586+00:00", "lang": "en", "topics": ["large-language-models", "ai-agents", "ai-safety", "ai-research"], "entities": ["Microsoft Research", "DELEGATE-52", "Gemini 3.1 Pro", "Claude 4.6 Opus", "GPT 5.4"], "alternates": {"html": "https://wpnews.pro/news/microsoft-research-llms-corrupt-your-files-during-delegated-work", "markdown": "https://wpnews.pro/news/microsoft-research-llms-corrupt-your-files-during-delegated-work.md", "text": "https://wpnews.pro/news/microsoft-research-llms-corrupt-your-files-during-delegated-work.txt", "jsonld": "https://wpnews.pro/news/microsoft-research-llms-corrupt-your-files-during-delegated-work.jsonld"}}