The AI vs AI Battle for the Internet

The episode covers Anthropic’s Claude Duet, Walmart’s willingness to let third-party agents shop, Meta’s new AI funding strategy, and a major cyber incident centered on the PHP2Shell vulnerability. For CIOs, the real takeaway is that agentic AI is moving into enterprise workflows and commerce while AI-assisted offense and defense are forcing faster patching, tighter key management, and better system visibility.

Key points

• Anthropic launched Claude Duet as a proactive, work-focused personal agent with explicit safety checks and a research-preview waitlist.
• Walmart will allow third-party agents to shop with linked accounts, spending limits, default approval, and purchase protections.
• Meta raised 2025 capex to $135B-$145B and created non-voting stock to fund AI infrastructure without diluting Zuckerberg’s control.
• A PHP vulnerability dubbed PHP2Shell could affect up to 70% of websites and requires urgent patching and incident response.
• Railway disclosed exposure of encrypted environment variables, reinforcing that AI infrastructure vendors are now prime targets.
• The broader shift is AI versus AI: defenders need agents to hunt, patch, and monitor faster than attackers can exploit.