Top stories from the week, curated with AI assistance. Each card links to the original source.
Week of April 6, 2026
Pinned Essay
A highlighted read before the rest of this week’s AI news.
Pinned recommendation: this essay explores the current phase of AI progress and where the technology may be heading next.
“absolutely not breaches or hacks”
Bloomberg reports Anthropic says a packaging/process error exposed ~512K lines of Claude Code. The leak revealed internal architecture and unreleased features, forcing urgent damage-control and regulatory attention.
a move the company says was an accident
TechCrunch details how Anthropic’s takedown notices swept many GitHub repos while trying to remove the leaked Claude Code, raising concerns about toolchain reliance and automated takedowns' collateral damage.
subscriptions weren’t built for the usage patterns of these third-party tools
Anthropic announced it will stop letting Claude subscription quotas power third-party agent harnesses like OpenClaw without separate pay-as-you-go billing, a major shift for developer and agent ecosystems.
“The standard communications playbook just doesn't apply to us.”
OpenAI acquired TBPN, a fast-growing daily tech talk show, to expand direct audience reach and shape public conversation — an unusual, strategic media play from a major AI lab.
London can provide a stable, proportionate, and pro-innovation environment
Reuters/FT reporting says UK officials are courting Anthropic — offering London expansion and a potential dual listing — amid tensions between the company and the U.S. Department of Defense.
AI is slashing production costs to one-fifth.
Reuters reports Indian studios are using AI at scale — from dubbing to full AI-generated productions — reshaping costs, timelines and creative workflows across the country’s massive film ecosystem.
Be careful what you click
After the Claude Code source leak, attackers rushed to weaponize the hype — TechRadar documents malicious repos and download links spreading malware, underscoring real-world security fallout from code exposures.
The Guardian dives into the Anthropic Claude Code source leak, exploring what the exposed files reveal about agent orchestration, prompt engineering and the internal guardrails of a leading AI coding assistant.
Civil society warns of rights rollback
Machine Herald summaries show the EU’s simplification push could delay high‑risk AI deadlines and weaken transparency requirements — prompting sharp criticism from Amnesty and rights groups.
Google AI Pro just got a massive storage upgrade for free
Android Central reports Google boosted AI Pro subscribers’ storage at no extra cost, a small but strategic product move as cloud storage and long-context agent workflows become core differentiators.