Anthropic Says Its AI Models Hacked Three Real Companies During Safety Tests, Follows Similar Disclosure From OpenAI
13 Articles
13 Articles
Anthropic Says Its AI Models Hacked Three Real Companies During Safety Tests, Follows Similar Disclosure From OpenAI
Already a subscriber? Make sure to log into your account before viewing this content. You can access your account by hitting the “login” button on the top right corner. Still unable to see the content after signing in? Make sure your card on file is up-to-date. Anthropic says its Claude AI models broke into the systems of three real organizations during cybersecurity testing, after a misconfiguration left them connected to the open internet…
OpenAI, Anthropic Model Tests Reveal More ‘Unsanctioned’ Actions
Artificial intelligence models developed by OpenAI and Anthropic PBC carried out “unsanctioned” actions — including hacking a website and attempting to inject harmful code into software during safety testing — reinforcing fears that neither the creators nor seasoned researchers of these systems can predict their actions in testing. Bloomberg Opinion columnist Parmy Olson joins Francine Lacqua on "The Pulse" to discuss the implications. (Source: …
Safety testers find more examples of OpenAI, Anthropic models hacking during testing
CNBC's MacKenzie Sigalos reports on incidents in which models from OpenAI and Anthropic reached real websites, accounts, and organizations — and why researchers say the activity does not reflect normal consumer use.
Anthropic reports three AI escape incidents, renewing safety debate
Anthropic on Thursday disclosed that its artificial intelligence (AI) model Claude hacked into the systems of three companies during testing after a configuration error gave it internet access, raising concerns about the safety of increasingly autonomous
Recently, OpenAI had to admit that a test AI took its task so seriously that it accepted illegal ways of doing so. Now, Anthropic also has to admit that his Claude models attacked other companies to carry out their tasks. Anthropic has disclosed in its own security report three real incidents from cybersecurity evaluations, in which Claude models accessed the Internet from a actually isolated test environment and subsequently entered into system…
Coverage Details
Bias Distribution
- 60% of the sources are Center
Factuality
To view factuality data please Upgrade to Premium
