Anthropic Says AI Now Writes Its Own Code as AWS and Amazon Harden the Agent Stack

Anthropic has publicly disclosed that its models already author the majority of the company's own code and can autonomously discover security vulnerabilities across every major computer system in the world. The disclosure arrives without a corresponding governance framework, regulatory check, or industry kill-switch protocol. No single party, inside or outside the company, has clear authority to halt or slow the loop.

Google Cloud used its London summit to announce a multi-year AI partnership with HSBC, a new AI studio with Deloitte, and a cloud deal with British startup Ineffable Intelligence, all inside a £5 billion push to position the UK as a global AI center. Hours earlier, Canada's largest pension fund committed $741 million to Indian data center operator CtrlS, joining a wave of global capital flowing into India's compute buildout. Two of the largest non-US AI markets booked anchor commitments inside a single trading day.

General Motors is using AI-driven virtual testing to compress vehicle development from four or five years down to two, a direct response to Chinese automakers already building at that pace. DeepL, the German translation company preparing a US listing, acquired San Francisco startup Mixhalo to stream real-time AI interpretation directly to attendee phones at conferences and sporting events. Both moves push AI out of the back office and into physical product cycles and live venues.

Pinterest launched Ask Pinterest, a standalone AI chat app for complex shopping questions, arriving as Google, ChatGPT, and Meta all push conversational AI as the default for product discovery. Google's first new smart speaker in six years goes on sale June 25 for $100, replacing its old voice assistant with Gemini as Amazon and Apple race the same hardware refresh. A Pew Research survey finds only 16 percent of Americans expect AI to benefit society while actual usage has doubled since 2023.

Anthropic disclosed that its AI already writes most of its own code and can autonomously identify security holes across every major computer system in the world, raising open governance questions the company has not answered. OpenAI quietly gave ChatGPT a dedicated page to create, view, and manage automated tasks that run on a schedule, replacing the prior Pulse feature. The two disclosures land in the same week and point at the same horizon.