Key artificial intelligence news in week 44 of 2026
The Swedish Government Offices restricted the use of the public web versions of ChatGPT and Claude on work computers because sensitive information could be entered into services lacking appropriate agreements and sufficient control over data processing. Google introduced Gemini 4 Carbon, a model intended for long-term programming and cybersecurity tasks, initially available only to selected partners. According to CrowdStrike, China-linked Vapor Panda used the ARTEX agent framework and the DeepSeek model to attack more than 200 organizations in at least 14 countries within 11 days and install the BLOODALCHEMY backdoor. Chainalysis reported that an attacker used attack methods based on AI agents to access data containing user IDs, names, email addresses and internal organizational roles; its research data, authentication data and session tokens reportedly did not leak. Canada plans to invest in Cohere and considers the development of secure and sovereign AI capability a national security issue.
CERT24