Given that Anthropic updates has been repeatedly referred to as one of the most safety-conscious companies in the AI industry, the revised policy represents a significant change.
Anthropic has updated its safety guidelines to better reflect the current global regulatory framework that places a higher priority on the development and competitiveness of AI.
The Claude manufacturer stated in an updated version of its Responsible Scaling Policy (RSP), a voluntary framework used by Anthropic to address catastrophic risks from AI systems, that it would continue to develop an AI model deemed dangerous even if a competitor had already released a model that was comparable or better.
This is different from their RSP from two years ago, which said that Anthropic will postpone potentially harmful AI development. Anthropic stated in a blog post on Tuesday, February 24 that the change in its safety policy was brought about by the rapid advancement of AI and the absence of government agreement on AI rules.
Given that Anthropic has been repeatedly referred to as one of the most safety-conscious companies in the AI industry, the revised policy represents a significant change. But the AI startup has also faced fierce competition from rivals like Google, OpenAI, and Elon Musk’s xAI, which frequently release state-of-the-art tools.
We believed that by revealing our RSP, other AI firms will be inspired to follow suit […]. We believed that RSPs and related principles would eventually be adopted as voluntary industry standards or inform AI regulations that promote safety and openness in the creation of AI models, according to Anthropic.
It further stated that “some parts of this theory of change have played out as we hoped, but others have not,” based on its evaluation of the earlier RSPs.
What is said in the new policy?
Anthropic emphasized three significant updates to its RSP. First, the corporation intends to keep its desired AI risk mitigations apart from the general AI safety recommendations it has given to global regulators and the industry. Second, Anthropic’s new RSP mandates that it create and publish a Frontier Safety Roadmap outlining its strategies for risk reduction in the areas of security, alignment, safeguards, and policy.
Lastly, the AI startup announced that third-party actors that are “deeply familiar with AI safety research, are incentivized to be open and honest about Anthropic’s safety position, and are free of major conflicts of interest” will be reviewing its Risk Reports.
US Pentagon versus Anthropic
The updated RSP coincides with growing hostilities between the US Department of Defense and Anthropic on limitations on the military applications of its Claude technologies.
According to Anthropic, company policies prohibit the use of its AI tools for autonomous violent acts or domestic surveillance. On Tuesday, February 24, however, US Defense Secretary Pete Hegseth informed Anthropic CEO Dario Amodei that the company had until the end of this week to loosen its usage guidelines.
Read more tech updates here
