EVA Tech BriefIT/DevOps

2026-06-10 - 7:30 PM PKT

Frontier LLMs

Anthropic Claude Fable 5 brings Mythos-class AI to public users

Anthropic's Claude Fable 5 launch is being framed as the first broad public release of its Mythos-class model family, with Mythos itself still restricted behind tighter access controls. The rollout is drawing attention not only for model capability, but also for the safety and usage-policy choices around high-risk cyber and bio domains.

GPUs and Silicon

Cloud and Infra

Pakistan Tech

TechJuice covers Anthropic's Fable 5 and restricted Mythos 5 launch

TechJuice reports that Anthropic launched Claude Fable 5 for general public use while keeping Claude Mythos 5 limited to selected cybersecurity and infrastructure partners. The story highlights June 9 availability and pricing, making the model release visible in Pakistan's tech news cycle.

Security

Microsoft patches exploited Exchange Server zero-day

Microsoft has patched an actively exploited Exchange Server flaw affecting Outlook Web Access users. The vulnerability allowed arbitrary JavaScript execution through cross-site scripting, so Exchange and OWA administrators should treat patching and exposure review as urgent.

The Gentlemen ransomware group rises quickly by offering affiliates 90 percent

Krebs profiles The Gentlemen, now described as the second most active ransomware group by victim count. The group's aggressive recruiting pitch, including a 90 percent affiliate share, shows how ransomware operations are competing for skilled operators and scaling quickly.

Dev Tools

Cohere open-sources coding agent that runs on one H100

Cohere has released North Mini Code as an open-source coding agent for teams building agentic development workflows. It runs on a single H100, which makes it a concrete alternative to fully managed coding models, though early testing points to higher output-token usage than comparable systems.

Azure API Management adds unified model API and MCP content safety

Azure API Management is adding a unified model API so clients can use one request format while APIM maps calls to Anthropic, Vertex AI, and other providers. Microsoft also expanded content-safety policies to cover MCP tool calls and agent-to-agent payloads, reflecting how governance is moving beyond plain LLM prompts.