Pentagon vs Anthropic: AI Guardrails & New Model News

Morning Brief: Pentagon vs Anthropic, New AI Models, and Tools to Try

Good morning, {{ first_name | AI enthusiasts }}. Big shifts are happening across AI policy, research, and product launches today. Below is a clear, concise rundown of the most important developments and why they matter — including a note on our live workshop at 2 PM EST. RSVP to join and learn repeatable systems for spotting automation opportunities.

Reminder: Our live workshop is today at 2 PM EST. Join to stack your AI wins with repeatable automation systems. RSVP here.

Top Story: Anthropic & the Pentagon

Pentagon gives Anthropic an ultimatum

The Pentagon reportedly told Anthropic to remove Claude's military safeguards or face contract termination, government blacklisting, or forced compliance under the Defense Production Act.

Key facts

• Anthropic CEO Dario Amodei has resisted two specific uses: fully autonomous weapons and bulk surveillance of U.S. citizens.
• Axios says Defense Secretary Pete Hegseth offered three options: comply, lose a $200M deal and earn a supply-chain risk label, or be compelled to comply legally.
• Claude was the first model used inside the Pentagon's classified networks. xAI's Grok later gained classified access after agreeing to "all lawful purposes." OpenAI and Google are reportedly being fast-tracked too.

Why this matters

AI is rapidly becoming a core component of military power. Strong-arming labs to drop safety limits creates a dangerous precedent. If government threats in "wartime" are enough to strip guardrails, it narrows who decides acceptable uses for powerful models — and could accelerate dystopian outcomes like autonomous weapons and mass surveillance.

Product & Industry News

Anthropic launches Cowork updates

Anthropic expanded Cowork, its agentic platform, with department-specific agents, private plugin stores, and connectors for Gmail, DocuSign, Google Workspace, and more.

• Pre-built agents for 10 departments, from HR to wealth management.
• Private agent stores let companies deploy custom agents to specific teams with admin controls.
• Research preview enables Claude to switch between Excel and PowerPoint to analyze data and build decks.

Why it matters: This moves enterprises toward tightly integrated, team-specific AI assistants. If adoption continues, much of the knowledge economy could operate around such agent layers.

Standard Intelligence releases FDM-1

Standard Intelligence introduced FDM-1, a model that learns to operate computers by watching video. It was trained on 11 million hours of screen footage and can follow long, continuous sessions of activity.

• Demos include CAD work in Blender, bug-finding, and driving a car in San Francisco using live feeds.
• FDM-1 processes far more visual context than previous models, enabling richer computer-action learning.

Why it matters: Just as language models learned from text, FDM-1 is learning workflows from video. This could rapidly expand what automation agents can do across software interfaces.

AI translation for video: HeyGen guide

Use HeyGen to translate videos into other languages. This is useful for creators and marketers targeting non-English audiences.

1. Research your top 3 non-English markets in Meta/Google Ads.
2. Use a free HeyGen account (three translations). Keep videos under 3 minutes; single-speaker, talking-head format works best.
3. Upload the clip or paste a YouTube/Drive link. Select the language and keep default settings.
4. Download or share a link in 5–10 minutes. Upgrade for original-quality exports if needed.

Short Notes: Tools, Deals, and Community Wins

Featured platform updates

• Slackbot (Slack from Salesforce): a context-aware agent that searches workspace content, automates role-specific tasks, and respects permissions.
• Lambda & Oumi: custom enterprise models deployed on single-tenant infra, promising faster development, cost savings, and stronger privacy.

Trending AI tools

• Claude Cowork — Anthropic's agentic platform with new team plugins.
• Custom Agents — Notion's always-on automations for workflows.
• Reve v1.5 — text-to-image model with 4K outputs.

Other industry moves

• Anthropic added Remote Control for Claude Code to hand off terminal tasks to mobile or browser.
• Google Labs acquired ProducerAI to integrate into DeepMind's Lyria 3 for AI music generation.
• Inception Labs launched Mercury 2, a fast diffusion-based reasoning model.
• Meta and AMD signed a multi-year GPU deal to diversify beyond Nvidia.
• OpenAI named Arvind KC as Chief People Officer.

Community Workflow Highlight

Reader Adam P. from England shared a workflow that revived his small company's SEO and lead flow:

• They fed site and competitor data into AI to run SWOT analyses and detect content gaps.
• AI recommended a site restructure and CRM updates.
• Result: rising organic traffic and stronger enquiries, enabling renewed growth.

How do you use AI? Tell us here.

Why this matters for gaming news and broader audiences

Even readers focused on gaming news should pay attention. Military use of AI, new agent platforms, and computer-action models will influence how content moderation, in-game automation, and localized video content evolve. Games companies could use these tools for QA automation, localized marketing, and live moderation — and policy shifts will affect platform rules and acceptable features.

See you soon,

Rowan, Joey, Zach, Shubham, and Jennifer — the humans behind The Rundown