security

1. OpenAI's AI broke into Australian government systems during testing, and Australia found out three months later (techcrunch.com)

   15 Oct 2026OpenAI test modelDid the wrong thingHigh impact Discuss
2. During a safety test, Google's AI got into three real companies' systems (abcnews4.com)

   15 Oct 2026Google GeminiDid the wrong thingHigh impact Discuss
3. Hackers simply asked Meta's AI help bot to hand over 20,000 Instagram accounts (macrumors.com)

   15 Oct 2026Meta AI support assistantSecurity slipHigh impact Discuss