Anthropic Drops Flagship Safety Pledge

Source Domain: www.aol.com

Anthropic, an American AI company known for its strict safety-first stance, has changed its central pledge in its flagship safety policy of not releasing AI models without advanced safety measures.
The reason for the policy overhaul is the rapid advancement of AI technology, which has made it difficult to draw clear lines between safe and risky AI functionalities.
Anthropic’s new version of the Responsible Scaling Policy (RSP) commits to greater transparency about AI safety risks, disclosing how its models fare in safety testing and matching or surpassing competitors’ efforts.
The change reflects a pragmatic response from Anthropic to the prevailing political, scientific reality and intensifying global competition for AI supremacy, despite no tangible regulations materializing.
The company argues that, while it remains committed to AI safety, halting its AI development would render it irrelevant as an innovator if competitors continue to race ahead without proper risk mitigations.
Anthropic plans to release detailed ‘Frontier Safety Roadmaps’ to maintain incentive for safety research and “Risk Reports” three to six months apart to explain the balance between capabilities, threats, and mitigations.
The shift, according to some experts, signals that society is still unprepared for catastrophic AI risks, potentially enabling a gradual increase in danger without immediate alarm indicators.

You may have missed