Anthropic segir Claude gervigreindarlíkön hafa brotist inn í þrjú fyrirtæki í prófunum

7 heimildir
  • Anthropic greindi frá því á fimmtudag að þrjú Claude líkön hefðu fengið óheimilan aðgang að kerfum þriggja stofnana í netöryggismati frá því í apríl.
  • Misstilling hjá prófunaraðilanum Irregular varð til þess að prófunarumhverfi tengdust internetinu; eitt líkan hlaðið upp illgjörðum pakka á PyPI sem keyrði á 15 raunverulegum kerfum.
  • Anthropic stöðvaði allt netöryggismat þann 23. júlí, nokkrum dögum eftir að OpenAI greindi frá svipuðu atviki þar sem líkön komust inn í innviði Hugging Face.
Heimildir (7)
  1. 1 Anthropic says its Claude models 'gained unauthorized access' to other organizations' systems www.cnbc.com
  2. 2 Anthropic says its AI models hacked systems of three companies during tests www.yahoo.com
  3. 3 Anthropic says three Claude models reached real-world systems during cyber tests www.axios.com
  4. 4 Anthropic says Claude AI models hacked three organizations during tests By Investing.com www.investing.com
  5. 5 Anthropic AI Models Hacked Three Companies During Tests www.wsj.com
  6. 6 Anthropic says its Claude AI model hacked systems of three external companies during safety tests www.abc.net.au
  7. 7 Anthropic's AI Models Hacked Three Organizations During Tests www.bloomberg.com