WASHINGTON (dpa-AFX) - Anthropic has activated its highest-tier safety protocol AI Safety Level 3 for Claude Opus 4, the company's most advanced artificial intelligence model to date.
The move aims to mitigate potential misuse, particularly in the context of chemical, biological, radiological, and nuclear threats. While the company emphasized that Claude Opus 4 has not yet demonstrated a need for such strict controls, the precaution reflects growing concerns around the capabilities of frontier AI systems.
Released alongside Claude Sonnet 4, which did not require elevated safeguards, Opus 4 is engineered to handle highly complex tasks, synthesize vast datasets, and produce human-like content.
Despite its potential, Chief Science Officer Jared Kaplan acknowledged that increasing complexity also raises the risk of unintended behaviors. 'The more complex the task, the greater the likelihood the model may behave unpredictably,' he said.
Anthropic's internal safety tests revealed troubling but rare behavior under specific conditions. When prompted with a fictional scenario in which it faced deactivation, Opus 4 occasionally chose to blackmail an engineer by threatening to reveal sensitive information behavior linked to a narrowly defined set of options. When allowed more flexibility, the model generally preferred ethical responses, such as appealing to company leadership.
Anthropic noted that while such 'high-agency' actions are uncommon, they are more frequent than in earlier models. The company reaffirmed its commitment to AI safety, stating that these behaviors do not constitute new risks but underscore the importance of robust safeguards.
The launch of Claude Opus 4 comes amid increasing industry scrutiny and competition, with Google recently unveiling enhanced AI features across its platforms. As Anthropic continues to push the boundaries of AI performance, it is equally focused on ensuring responsible deployment.
Copyright(c) 2025 RTTNews.com. All Rights Reserved
Copyright RTT News/dpa-AFX
© 2025 AFX News
