AI DAILY / DEV
←
THURSDAY
June 4, 2026
→

    Anthropic Maps a Year of AI Cyber Misuse to MITRE ATT&CK — 67% of Banned Actors Used Claude for Malware
    • Anthropic studied 832 accounts banned for malicious cyber activity between March 2025 and March 2026, mapping 13,873 actions across 482 unique ATT&CK techniques.
    • Medium- or high-risk actors using AI for cyber ops jumped from 33% to 56% across the year — a 1.7× rise the report frames as a structural shift, not a blip.
    • Malware development dominates: 67.3% of the 832 actors used Claude to build malicious software; only 6.5% reached more advanced 'lateral movement' help.
    • Subset of findings ships in Verizon's 2026 DBIR; Anthropic also publishes an interactive LLM ATT&CK Navigator at red.anthropic.com.
    research anthropic.com

    Microsoft Ships MAI-Code-1-Flash and MAI-Thinking-1 — First In-House Models for Copilot, Claim Lead Over Claude Haiku 4.5
    • June 2 Build keynote: MAI-Code-1-Flash is a 5B-parameter coding model rolling out to all GitHub Copilot tiers (Free → Max) starting in VS Code.
    • Microsoft reports 51.2% on SWE-Bench Pro vs Claude Haiku 4.5's 35.2%, with up to 60% fewer tokens on harder Verified problems via 'adaptive solution length'.
    • MAI-Thinking-1 is a 35B-active-parameter sparse MoE reasoner with 256K context, trained entirely on commercially licensed data (no third-party distillation), in private preview on Foundry.
    • HN thread headlines the production-harness training story — model was tuned inside the live Copilot loop, not benchmarked into it after the fact.
    models microsoft.ai

    xAI Makes Grok the Default Voice for 2.5M Vapi Agents
    • June 3 — Grok Voice becomes the engine behind all 12 of Vapi's core voices, the speech layer for 2.5M+ voice agents already deployed on the platform.
    • Vapi's blind eval ranked Grok Voice #1 against incumbent providers; a 4,500-vote X poll split 50/50 trying to distinguish a Grok clone from the human original.
    • xAI's TTS and low-latency STT APIs are now reachable through the Vapi SDK, pricing positioned directly against ElevenLabs (TTS) and OpenAI Whisper (STT).
    • Frames Grok Voice as the cost/quality leader after last week's Voice Agent API debut — xAI now owns the model AND the distribution channel.
    tools x.ai

    OpenAI Pitches a Federal Blueprint for Frontier AI — Mandatory CAISI Evals, No Pre-Approval
    • June 3 policy paper proposes a three-part U.S. framework: a single national standard built on state frontier-safety laws, CAISI as the lead federal evaluator, and a cross-agency resilience plan.
    • CAISI would run mandatory pre-deployment evaluations of 'the most capable frontier models' and monitor progress toward recursive self-improvement — but only recommend, never block or license.
    • Lands one day after Trump's voluntary 30-day disclosure EO; OpenAI's pitch is essentially the durable statute version of that order.
    • Pairs with a separate June 3 brief on youth safety and OpenAI's election-safeguards 2026 update — coordinated policy push, not a one-off.
    industry openai.com

    Anthropic Launches Services Track for Claude Partner Network — 40,000 Firms Applied Since March
    • Three-tier program (Select, Preferred, Global Premier) gates partner status on certified practitioners, production deployments, and public customer endorsements.
    • Global Premier requires 1,000 certified practitioners, 100 customers across three+ regions, 15 public endorsements, and a joint business plan with Anthropic.
    • Anthropic says 40,000+ firms have applied since the March launch and 10,000+ consultants have earned a Claude certification.
    • Companion Claude Partner Hub portal refreshes partner standings daily — pitched as Anthropic's enterprise-channel answer to AWS/Google/Microsoft partner programs.
    industry anthropic.com