Anthropic has updated its usage policy to reflect growing concerns about how artificial intelligence can be misused in today’s volatile landscape. The company’s new rules, effective mid-September, focus on cybersecurity threats, political manipulation, and the development of dangerous weapons. The changes are part of a broader effort to keep AI applications aligned with safety principles.
One of the most notable clarifications centers on cybersecurity. Anthropic now explicitly forbids using tools like Claude Code and Computer Use to create malware, compromise networks, or launch denial-of-service attacks. By drawing these sharper boundaries, the company is aiming to prevent its models from being misused by hackers or those seeking to disrupt online systems.
Restrictions on weapons and hazardous materials have also been reinforced. Anthropic has expanded its earlier general ban to directly prohibit assistance in creating high-yield explosives or chemical, biological, radiological, and nuclear weapons. This move reflects heightened concerns about AI being exploited in contexts with potentially catastrophic consequences.
The policy also introduces a more nuanced approach to political content. Instead of blocking all campaign-related uses, Anthropic is focusing on preventing manipulative practices that could undermine democratic processes. These updated rules apply primarily to consumer contexts, while enterprise use remains under a different framework. The company has paired these steps with its AI Safety Level 3 standards to reinforce responsible scaling.
#Anthropic #AI #ArtificialIntelligence #Cybersecurity #AIethics #TechPolicy #ClaudeAI #ResponsibleAI #AInews