Anthropic Adjusts AI Safety Policy Amid Competition
Anthropic, a company started by former OpenAI staff, has recently shifted its approach to artificial intelligence safety. Originally built on the idea of developing AI with strong safety practices, the company has now made significant changes to its policy regarding how it handles AI advancement.
On Tuesday, Anthropic announced the release of the third version of its Responsible Scaling Policy (RSP). This framework was designed to help prevent serious risks associated with AI systems. However, the company revealed that due to increased competition and a lack of government regulations, it would no longer pause or delay the development of new models when safety measures cannot keep up.
Previously, Anthropic’s safety policy required a stop on training more powerful models if those models surpassed the company’s ability to manage their risks. This safeguard has now been removed.
In explaining the change, Anthropic pointed out that the landscape surrounding AI has shifted. Currently, there is more focus on competitiveness and economic benefits than on safety discussions, particularly at the federal level.
Jared Kaplan, Anthropic’s chief science officer, shared with Time Magazine that the original safety policy was falling behind in the fast-paced AI race. He expressed that it wouldn’t benefit anyone for the company to halt training while competitors were advancing rapidly.
While Anthropic continues to advocate for safety measures within the AI industry, it plans to separate its own strategies from its recommendations for others. The company believes effective government collaboration on AI safety is essential, but acknowledges that this will require time and effort.
This policy shift comes at a delicate moment for Anthropic, as it is currently in a dispute with the Pentagon. Recently, Defense Secretary Pete Hegseth gave CEO Dario Amodei a deadline to roll back AI safety measures or risk losing a significant $200 million defense contract. Despite this pressure, sources indicate that the change in Anthropic’s safety policy is not directly linked to the Pentagon situation.
