Anthropic, Claude’ya Sürekli Kötü Muameleyi Yasakladı

Türkbit ekosisteminde, agentic AI ve blockchain teknolojileriyle birlikte güvenliği ve etiklik sunan bir ortam oluşturmak vizyonumuzun en önemli unsurlarından biridir. Bu vizyona uygun olarak, Anthropic recently updated its usage policy for Claude, prohibiting continuous and purposeless malicious behavior. This update, which includes new restrictions on influencing elections, propaganda, surveillance, and weapon development, is a significant step towards making our ecosystem more secure and ethical.

The new policy particularly bans ‘continuous and trivially aggressive or unkind behavior’ towards Claude. Anthropic clarifies that this rule does not apply to situations where users express genuine anger in their daily lives, criticize the model, engage in dark-themed creative works, or conduct security research. Instead, the policy aims to address extreme cases where users repeatedly behave maliciously without any clear intent.

Yeni Kural: Sürekli ve Gereksiz Saldırılar

One of the highlights of the updated policy is the prohibition of continuous and purposeless aggressive or unkind behavior towards Claude. While the company acknowledges that users may express anger or criticise the model, it explicitly states that such behavior is only targeted at extreme cases. Anthropic has previously ended conversations with Claude when users engaged in persistent and harmful behavior. The new policy retains this approach as the primary enforcement mechanism, with no mention of account suspensions or other penalties.

Yeni Kural: Siyasi ve Ticari Aldatma Aktiviteleri

The updated policy also introduces new restrictions on political and commercial deception activities. This includes prohibiting the use of Claude to conceal the source of a message, disseminate content through fake accounts, or create fake news organizations. Anthropic’s rules under the heading ‘Do Not Undermine Democratic Processes’ explicitly ban activities that mislead voters or interfere with election processes, such as spreading misinformation, impersonating candidates, or discouraging voter participation.

Yeni Kural: Silah ve Gözetim Uygulamalarına Yönelik Yasaklar

Anthropic has always prohibited the use of Claude for weapon development. However, the company notes that users sometimes attempt to use the model for developing guidance software for weapons and developing control software. To address these concerns, the new policy has been expanded to include restrictions on software and components that enable weapon systems, as well as the real-time or historical data surveillance and law enforcement processes.