KI: Anthropic dämmt heimliche KI-Kopien durch Konkurrenten ein
Anthropic hat den Zugriff auf geschützte Denkprozesse über Proxy-Netzwerke erschwert. Das Unternehmen hatte zuvor auch chinesische Start-ups des Abgreifens von Daten beschuldigt.
Anthropic, a US-based artificial intelligence company, has significantly curbed the unauthorized data harvesting by competitors, according to the company's own statements. The new protective measures for their in-house Claude models have reduced the amount of data stolen during so-called "distillation attacks" by 90 to 99 percent, the company reported on Monday.
In this method, rivals attempt to copy the capabilities of advanced artificial intelligence to develop their own models more quickly and cost-effectively. Anthropic has now made it more difficult for protected thinking processes of Claude to be accessed through proxy networks and has blocked tens of thousands of accounts. The company cited the Chinese AI startup Z.ai (Zhipu) as an example, which had tried to copy capabilities of Claude to develop its own model GLM-5.3.
Such practices posed a risk to national security, Anthropic explained, amid increasing technological competition between the US and China. Z.ai initially did not respond to a request for comment from the news agency Reuters.
Written by urgent.news from Handelsblatt's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
