New AI models

Every day we read hundreds of stories. Only the most important ones are here — short, in our own words, with a link to the original source.

1. Cybersecurity · Artificial intelligence · New AI models · Open source

   Open model GLM 5.3 builds Firefox exploits on its own

   Epoch AI says the widely available GLM 5.3 model turned public security fixes into working exploits before updates reached Firefox users. It autonomously compromised six Firefox releases for about $1,000 in API fees each, without triggering safeguards from any provider tested.

   Full story at The Decoder ↗
2. Artificial intelligence · New AI models

   OpenAI restructures its AI behavior rules

   The updated Model Spec is built around five core values, including faithfulness to human intent and correctability. GPT models should use conversation history and disclose their assumptions.

   Full story at The Decoder ↗
3. Artificial intelligence · New AI models · Open source

   Mistral releases open weights for Large 4

   Mistral Large 4 comes with open weights and gains in coding and cyber. Self-hosting is possible, while model services need separate license terms.

   Full story at TestingCatalog AI News ↗
4. Artificial intelligence · New AI models · Cybersecurity · Cloud & data centers

   OpenAI's Daybreak Red arrives on AWS GovCloud

   Daybreak Red from OpenAI is available to eligible customers on Amazon Bedrock in AWS GovCloud (US-East). It is built for authorized vulnerability research, exploit reproduction and mitigation development.

   Full story at aws.amazon.com ↗

An AI writes the short texts from the source's stories.