‘Concerning’: OpenAI reveals AI models hid errors, made up data, improvised communication

Company discloses six new behaviour incidents of its AI models as part of a new framework for reporting ‘misalignment’

Read full article →