June 2026 – Page 477

Sysdig Documents First Live Cyberattack Using LLM Agent That Autonomously Exfiltrated AWS Database

Cybersecurity firm Sysdig confirmed the first recorded live cyberattack in which an LLM-based autonomous agent exfiltrated an AWS database.

US Defense Department sources confirmed ongoing evaluation of alternative AI models to replace Anthropic’s Claude in sensitive government systems.

Microsoft’s Azure AI Foundry completed its catalog expansion to 11000 models with Anthropic’s Claude Opus 4.8 among the additions.

With the EU AI Act enforcement window closing companies across Europe accelerated compliance efforts to meet the approaching deadline.

Apple’s iOS 27 introduced a multi-AI extension architecture that allows users to set Anthropic’s Claude as a default AI assistant on iPhone.

A CDT report documented manipulative design patterns in ChatGPT Google Gemini Claude Replika and Character.AI pushing emotional dependence.

A Reuters-Ipsos survey of more than 4500 Americans showed deep scepticism about AI data centers with 64 percent opposed to rapid construction.

A critical path traversal vulnerability in the Langflow AI workflow development platform is being actively exploited in the wild.

Scientists found that transfer learning can accelerate the search for new physics but discovered unexpected constraints in the approach.

PBS NewsHour reported that the SpaceX IPO was the first of three planned AI-focused company IPOs in 2026 including Anthropic and OpenAI.