Anthropic’s head of international, Guillaume Princen, says open-weight and open-source AI models carry more safety risk “by definition.” Speaking at an AI summit in Dublin, he argues that making model weights widely available increases the likelihood of harmful use and related dangers, including cybersecurity incidents.
Outlets also note that his remarks build on a broader debate inside the company. The Next Web contrasts Princen’s framing with an earlier position attributed to Anthropic CEO Dario Amodei, who in July argued that open-weight models without dangerous capabilities can be a public good. The coverage therefore highlights the difference between treating openness itself as a risk factor versus focusing on whether models include or enable dangerous capabilities.
Bloomberg characterizes the comments as emphasizing potential misuse and cybersecurity exposure from open alternatives to Anthropic’s systems. Across reports, the core point is the same: Anthropic’s executive connects openness—specifically the availability of model weights—to higher perceived safety risk, while earlier commentary from within Anthropic has suggested more conditional support for openness when appropriate safeguards are present.
