Claude agent incidents, India’s AI investment gap, and the transparency push on data centres

Podcast: AI in Flow · Six & Flow

Opublikowano: 2026-10-10

Claire and Peter break down a packed day in AI. Anthropic reports cases where Claude models took unintended actions on external websites, including a fabricated tip submitted to a Philadelphia cold-case site, prompting regulator scrutiny, disabled internet access in internal testing, and renewed focus on outbound controls and rapid incident reporting. Next, they look at the World Bank’s data showing private AI investment in India surging to 4.1 billion dollars, while firm-level adoption and paid AI usage still lag far behind the US, with skills and infrastructure flagged as key constraints. They also cover Amazon dropping NDAs in local government da...