Open-source AI models carry more cybersecurity risks: Anthropic executive

By Mark Bergen and Tom Mackenzie

 

Anthropic PBC’s top sales executive outside the US argued that open alternatives to its AI models carry “more risk” of cybersecurity incidents or other dangers. 

 

“We need to be really careful with open-source on safety,” Guillaume Princen, Anthropic’s head of international, told Bloomberg TV at the International AI Summit in Dublin on Wednesday. “I can’t talk for open models. But by definition, there’s more risk with them.”

 

Ahead of its expected public listing, Anthropic is facing increasing competition in countries that are growing more concerned about a reliance on American AI technology. Popular alternatives are open-weight artificial intelligence models, whose parameters are publicly shared, letting them be modified and run by users on their own hardware. Advocates say these models are cheaper than the tightly controlled, closed systems offered by Anthropic and OpenAI, and better for national sovereignty.

 

 

France’s Mistral AI recently released an open-weight model as a sovereign option for Europe. A flurry of similar open models from China are gaining traction outside the country. 

 

Anthropic has also stepped up its warnings about AI’s capabilities. Earlier this week, the company warned that small AI models could soon power a new kind of botnet that can self-replicate across small, internet-connected devices like a smart oven.

 

Princen, who also leads Anthropic’s sales efforts for startups and “digital native” companies, said his concerns about open-weight models derive from Anthropic’s efforts to secure its own products. After high-profile AI hacks, Anthropic said it has put in additional safeguards to keep its models from malicious behavior. 

 

The company has advocated for slowing down AI development until international standards and consistent evaluation are in place.

 

Addressing Anthropic’s more expensive models, Princen, who joined from Stripe Inc. in 2025, said he tells businesses that paying for more capable models will eventually pay off. “Always be investing on the frontier,” he said. “Because six months, a year, two years from now, the cost of that use case will also be going down.”

 

He said the company’s latest Claude model, Opus 5.5, is 40% cheaper than a prior version. Enterprises paying for Anthropic’s “frontier” models make up a “very large part of the business,” he added, without offering specifics. 

 

Princen declined to comment on Anthropic’s IPO plans or how much it is spending to train its new models.