Watchtower tracks changes to corporate and government AI safety policies, both announced and unannounced. Click any entry for details.
Aug 8, 2024
OpenAI
Moderate
On August 8, 2024, OpenAI released their promised preparedness scorecard for GPT-4o. This scorecard grades GPT-4o on multiple risk categories including cybersecurity, biological threats, persuasion, and model autonomy.
In July, The Midas Project reached out to OpenAI to inquire about why a preparedness scorecard had not yet been released for GPT-4o, even after the model had been released. They did not return our message.
The preparedness scorecard, as well as the full GPT-4o system card, can be found here.