AI digest: Industry upheaval and technical breakthroughs
OpenAI faces internal rebellion over Pentagon deals while researchers push forward with new AI architectures and enterprise applications scale rapidly.
A week of corporate drama and genuine technical progress. The Pentagon’s AI partnerships are causing more fallout than expected.
OpenAI robotics chief quits over Pentagon deal
Caitlin Kalinowski has resigned from OpenAI in protest of the company’s Defence Department agreement. This follows growing internal opposition to military applications of AI. The departure of a key robotics leader suggests OpenAI’s pivot towards government contracts is costing them talent in critical areas.
LeCun proposes ditching AGI for “Superhuman Adaptable Intelligence”
Yann LeCun’s team argues that AGI is poorly defined and proposes Superhuman Adaptable Intelligence (SAI) as a more precise goal. SAI focuses on systems that can rapidly adapt to new domains rather than matching human-level performance across all tasks. This isn’t just academic hairsplitting - clearer definitions could reshape how we build and evaluate AI systems.
Claude finds over 100 Firefox vulnerabilities
Anthropic’s Claude has discovered more than 100 security bugs in Mozilla’s Firefox browser, including vulnerabilities that human testers missed for years. This demonstrates AI’s potential for automated security auditing at scale. If AI can systematically find bugs faster than traditional methods, it changes how we approach software security entirely.
Enterprise AI spending accelerates
JPMorgan is pushing its tech budget toward £15.8 billion in 2026, with AI investment driving much of the increase. Meanwhile, Dyna.Ai raised an eight-figure Series A to deploy agentic AI in financial services. The shift from pilot projects to production systems is finally happening at enterprise scale.