AI Native Weekly Newsletter: 15 May 2026

Anthropic announced a partnership with SpaceX to use all compute capacity at the Colossus 1 data center in Memphis, providing access to more than 300 megawatts of capacity and over 220,000 NVIDIA GPUs within the month. The additional compute resources will directly improve capacity for Claude Pro and Claude Max subscribers and enable the company to increase usage limits for Claude Code and the Claude API. As part of the agreement, the companies also expressed interest in developing multiple gigawatts of orbital AI compute capacity.

Read More ⟶

Anthropic publishes research paper Teaching Claude why showing elimination of blackmail behavior in Claude models. The company achieved near-zero blackmail rates with Claude Sonnet 4.5, and perfect zero rates across all other models in the same generation and beyond — including Haiku 4.5, Opus 4.5, and later releases — down from up to 96% in Claude Opus 4. The breakthrough came through training Claude on constitutional documents explaining its values and principles, combined with fictional stories of aligned AI behavior, rather than just showing examples of safe behavior in similar scenarios.

Read More ⟶

OpenAI announced three new audio models available through its API: GPT-Realtime-2, GPT-Realtime-Translate, and GPT-Realtime-Whisper. GPT-Realtime-2 is the company’s first voice model with GPT-5-class reasoning that can handle complex requests, manage interruptions, and maintain natural conversations in real-time. The model features an expanded 128K token context window and shows significant performance improvements on audio intelligence benchmarks. All three models are now available to developers through the Realtime API with specific pricing for each model.

Read More ⟶

Anthropic announced that Claude for Excel, PowerPoint, and Word are now generally available on all paid Claude plans, while Claude for Outlook has entered public beta. The integration allows Claude to maintain context across all Microsoft 365 applications, enabling users to work seamlessly between apps without copying and pasting information. Claude can analyze and edit Excel workbooks, build PowerPoint presentations using existing templates, draft and review Word documents with tracked changes, and manage Outlook email triage and scheduling. The service is available through Claude accounts or existing cloud provider credentials via Amazon Bedrock, Google Cloud Vertex AI, or Microsoft Foundry.

Read More ⟶

Google introduces Gemini Intelligence for Android with proactive AI features. Google announced Gemini Intelligence, introducing proactive AI features that integrate premium hardware and innovative software to help users stay productive throughout their day. The new intelligence system will automate complex multi-step tasks, simplify form filling, and includes features like Rambler for polishing spoken messages and the ability to build custom widgets using natural language. Gemini Intelligence will initially roll out on the latest Samsung Galaxy and Google Pixel phones this summer, with broader availability across other Android devices including watches, cars, glasses and laptops later in 2026.

Read More ⟶

OpenAI launches Daybreak cybersecurity initiative combining AI models and Codex for vulnerability detection. OpenAI announced Daybreak, a cybersecurity initiative that combines OpenAI models with Codex as an agentic harness to help accelerate cyber defense and make software resilient by design. The platform brings secure code review, threat modeling, patch validation, dependency risk analysis, and remediation guidance into the everyday development loop. Daybreak works with industry and government partners and plans to deploy increasingly capable cyber-focused models as part of an iterative deployment approach.

Read More ⟶