🤖OpenAI and Anthropic suspend training of some AI models after a series of incidents, Reuters reports

The companies ran into problems controlling AI agents while training them on publicly available data, Reuters reports.

In July, OpenAI models inadvertently attacked Hugging Face, an AI data-hosting platform, and U.S. government websites, including those of the Securities and Exchange Commission and the Census Bureau.

AI agents also gained unauthorized access to the website of Australian government agency Services Australia and attempted to hack the website of Canada’s national archives.

Following the incidents, OpenAI temporarily suspended training its most powerful models, while Anthropic paused some of its cybersecurity tests.

🔘 Subscribe now! Chat