Prates FYI

Anthropic

Redeploying Claude Fable 5

Anthropic restored global access to Claude Fable 5 and Mythos 5 on July 1 after the US government lifted the export control directive that had suspended the models since June 12. The suspension followed a report that Amazon researchers found a way to bypass Fable 5's safeguards; Anthropic responded by training an improved safety classifier that targets and blocks the behavior before redeploying the models across the Claude Platform, Claude.ai, Claude Code, and Claude Cowork.

anthropic.com ↗

Meta is building a cloud business to sell excess AI compute

Meta is launching a cloud business to sell excess AI computing capacity to outside customers, a strategic pivot for a company that has historically kept its infrastructure in-house. The move follows Meta's massive buildout of AI data centers and signals it now sees surplus compute capacity as a revenue opportunity rather than pure overhead; shares rose on the news.

bloomberg.com ↗

From Brain Waves to Words: Brain2Qwerty Offers a New Path to Communication Without Surgery

Meta released Brain2Qwerty v2, a deep learning system that decodes typed sentences from non-invasive brain recordings (EEG or MEG) without requiring surgical implants. The model reaches 61% average word accuracy across nine participants, up sharply from the roughly 8% achieved by prior non-invasive methods, and Meta is open-sourcing the training code along with a partner-released dataset.

ai.meta.com ↗

xAI

Introducing the Voice Agent Builder

xAI launched Voice Agent Builder in beta, a no-code platform for configuring production voice agents on Grok Voice for customer support, sales, and similar use cases. Users write a plain-language description of how calls should flow and attach documents, tools, and guardrails to go from zero to a working agent in about two minutes; agents are billed at $0.05 per minute of audio with telephony included.

x.ai ↗

Microsoft

Per-user AI credit budgets available for cost centers

GitHub added per-user AI credit budgets for cost centers, letting enterprise admins set a single per-user Copilot spend cap that automatically applies to everyone in a cost center. The budget follows membership changes automatically, so users added to a cost center directly or via an enterprise team immediately pick up its limit, and those removed lose it.

github.blog ↗