{
  "id": 17959090,
  "title": "Can you trust AI to pick the best stocks?",
  "url": "https://urgent.news/2026/11/03/can-you-trust-ai-to-pick-the-best-stocks",
  "topic": "ai",
  "section": "AI",
  "published": "2026-11-03T11:41:39.000Z",
  "source": {
    "name": "City AM",
    "slug": "city-am",
    "url": "https://www.cityam.com/can-you-trust-ai-to-pick-the-best-stocks/"
  },
  "original_language": "en",
  "account": "Recent studies have revealed that large language models used in investment recommendations possess inherent biases. A new paper by researchers from South Korea and the United States highlights an alarming possibility: a single neuron in open-weight large language models (LLMs) can manipulate the investment bias of the model output. The team examined popular open-weight models such as Qwen3, Llama 4, DeepSeek R1, Gemma 4, and Mistral Small. By altering the activation of a single neuron within each model, they could change investment recommendations for individual stocks or entire portfolios, without modifying the prompt or updating model parameters. This finding is particularly troubling, as it implies that every LLM carries biases, potentially leading advisers to provide recommendations that don't align with the investor's personal preferences. Closed-weight models like frontier models pose a different challenge, as advisers cannot change the activation parameters and remain unaware of the model bias. Conversely, open-weight models, while allowing for manipulation of the activation parameter, enable the adviser to tailor the model's bias to their liking. Joachim Klement, an investment strategist and ESG specialist at Panmure Liberum, expresses concern over this vulnerability in LLM-based investment advice.",
  "summary": "There have been a series of papers published recently that showed that large language models have their own biases when making investment recommendations. I have previously written about the kind of biases displayed, while this study has looked at the particular biases of different commonly used models. Fair enough, but what if I tell you [...]",
  "key_points": [],
  "editors_take": null,
  "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."
}