{
  "id": 20355766,
  "title": "AI could help speed up research ethics approvals",
  "url": "https://urgent.news/2026/11/17/ai-could-help-speed-up-research-ethics-approvals",
  "topic": "ai",
  "section": "AI",
  "published": "2026-11-17T00:40:04.000Z",
  "source": {
    "name": "Phys.org",
    "slug": "phys-org",
    "url": "https://phys.org/news/2026-11-ai-ethics.html"
  },
  "original_language": "en",
  "account": "A new study from the University of Bristol has found that large language models (LLMs) can significantly speed up research ethics approvals. Researchers tested three widely used AI models - Claude 3.5 Sonnet, ChatGPT-4o, and Microsoft Copilot - against published quality standards. The results showed that Claude was the most accurate, correctly identifying requirements 87.9% of the time, compared to 66.5% for ChatGPT and 62.8% for Copilot. Importantly, Claude was also the fastest, taking just over seven minutes on average to check each application, while the other models took between 13 and 23 minutes. The study highlights the potential for AI to assist in the early stages of the research approval process, by automating routine checks and freeing up staff time to focus on ethical decision-making. However, the researchers caution that while LLMs can identify missing information and potential deviations from standards, they should not replace human judgment. Lead author Dr. Panos Maghsoudlou notes that the findings suggest AI could \"help reduce the administrative burden\" associated with approvals, allowing staff to concentrate on substantive ethical issues. The study also found that the models made different types of errors, with Claude producing fewer \"hallucinations\" - false information. Despite these promising results, the researchers emphasize that further testing is needed before AI can be adopted to improve research approval processes.",
  "summary": "Artificial intelligence programs could be used to improve efficiency for professionals dealing with research ethics reviews, enabling humans to focus on the ethical decision-making.",
  "key_points": [
    "Claude 3.5 Sonnet AI model most accurate, identifying requirements 87.9% of the time",
    "Claude AI fastest, taking over 7 minutes per application on average",
    "AI models can reduce administrative burden, allowing focus on substantive ethical issues"
  ],
  "editors_take": "Automating routine administrative checks with large language models shifts the workload from clerical verification to substantive ethical decision-making, potentially reducing the time required for research approval processes.",
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}