Anthropic 发布威胁报告，披露 Claude 全球滥用案例

PLUS: How Rowan used Astra to redo his wardrobe

Zach Mink

Good morning, AI enthusiasts, and welcome to our 3,672 new readers. Need a feel-good AI palate cleanser after yesterday’s extinction doom? Anthropic has the opposite: a full accounting of everything bad actors attempted with Claude over the last eight months.

A 150+ page threat report walks through case studies of misuse it shut down across the globe, from rocket guidance, bioweapons, and espionage to Chinese labs training off Claude (and even serving the model to customers as their own). Feeling better yet?

In today’s AI rundown:

Anthropic opens the files on global Claude misuse

Rowan’s Corner: I used Astra to redo my wardrobe

A beginner’s guide to ChatGPT Work

DeepSeek turns up the pressure on AI pricing

LATEST DEVELOPMENTS

Image source: Anthropic

The Rundown: Anthropic published its latest Threat Report, detailing the most notable cases of Claude misuse it disrupted over the last eight months, including bioweapon flags, espionage, and acts of Chinese AI labs serving Claude in place of their own model.

The details:

Anthropic named seven Chinese labs behind distillation efforts, including Alibaba, DeepSeek, Moonshot, and Xiaomi, leaning on thousands of fraudulent accounts.

Moonshot and DeepSeek even served Claude to their customers in some instances and passed it off as their own model, then used responses for training.

Five biology cases from scientists using Claude raised flags for possible weapons applications, but Anthropic said it “does not assert that they intended harm.”

One operation in Yemen used Claude Code to build guidance software for a rocket, with the actor going back to Claude for advice after the test-flight failed.

Another consultant built Mali's spy agency a system aiming to watch 25M phone lines, with similar surveillance attempts banned in Iran, China, and more.

Why it matters: This is just a brief, so read on for Claude running 4,700 dating-app personas, cloning an activist's writing style to talk with his contacts, rebuilding malware to avoid antivirus detection, and more. These were all attempted with Opus-level models and below, making future reports with more capable models even harder to fathom.

The Rundown: You may have the best product, but no buyers will sign these days without proof of your security. Example: A prospect asks for proof of compliance. The deal stalls while you scramble, your engineer gets pulled off the roadmap to audit prep, and every enterprise conversation turns into a fire drill.

With Vanta, you’ll:

Get compliant fast—SOC 2, ISO 27001, HIPAA, and more

Stay compliant and build a strong security foundation with continuous control monitoring

Access the Vanta agent everywhere you work, even in Claude or Cursor

Vanta is trusted by more than 16,000 companies like Ramp, Harvey, and Writer. Learn more by watching their on-demand demo .

Rowan: OpenAI’s GPT-6 Astra model came out last week to ChatGPT paid plans, and I’ve been having an absolute blast pushing it to its limits (and waiting for limits to reset…).

The capabilities that make it genuinely feel like a step up are its computer use, visual consistency, long one-prompt jobs -- and, to my pleasant surprise, personal life use cases.

One personal use case that is improving my life is getting Astra to redesign my entire wardrobe and give me a system for picking clothes based on real-time local weather.

For quick context, I wear a lot of black, usually a black T-shirt and jeans. It works in tech, but I've been trying to level up my wardrobe. The problem is that my absolute worst nightmare is a closet full of clothes creating more clutter in my brain every day when picking an outfit.

So I systematized it (tech nerd brain, can't help it). One outfit per day of the week, per season, plus gym and lounge: 30 looks total, everything hung up and ready.

Astra turned that idea into an app. I gave it full-body portraits, my sizes, my colors, some niche details about my style, and a master prompt, and it generated 30 looks with 70 try-on images where I'm the model in every photo, pulling real-time local weather so the daily pick changes with what's outside.

The virtual try-on is genuinely incredible. If I can't find a piece online, I update the model, and it rebuilds the wardrobes.

I posted my exact step-by-step and full master prompt in our Workflow Hub . Hope it’s useful to those in a similar boat.

The Rundown: In this guide, you will learn how to set up a ChatGPT Work project for a recurring workflow, processing the work once and then automating it.

Step-by-step:

Create a local folder with a two-digit prefix. In ChatGPT desktop, open New chat > Work > Choose project > New project, name it, and select that folder

Tell ChatGPT about your project and ask it to propose a structure. We used it to process meeting notes and turn those into action items and weekly plans

Now add in some input documents and tell ChatGPT to process them once

Once it's done, tell it to create three useful automations that run on a schedule to automate this processing

Pro tip: Tell ChatGPT to turn those automated routines into skills that you can call at any time manually with a slash command.

The Rundown : Tines 3B gives every team the power to build with AI, and gives IT the control to run and monitor that work safely.

With Tines 3B, you can:

Turn ideas into production-ready apps, agents, and automations in any stack

Keep data secure at every step and ship workflows that improve over time

Get end-to-end visibility on spend, workflow performance, and value

Image source: DeepSeek

The Rundown: Chinese AI lab DeepSeek just released V4.1-Flash, an efficient new open-weight model that edges out systems like Claude Opus 5 and GPT-5.6 Sol on several agentic, coding, and cyber benchmarks at extremely low price points.

The details:

Flash runs at about a quarter of DeepSeek V4-Pro's per-token price and still outscores it across the board, now acting as the company’s default model.

V4.1 scored a 40 on AA’s Intelligence Index, still well behind the frontier, but showing strong results in agentic, coding, and cyber benchmarks.

Pricing comes in at just $0.15 / 0.60 per million tokens, with DeepSeek publishing the model’s weights on Hugging Face with an MIT License for download.

Flash joins Z.ai ’s GLM-5.3-Flash as the two strongest intelligence systems in their price range, with DeepSeek still set to release larger models in the new family.

Why it matters: DeepSeek's efficiency and pricing remain its edge, with Chinese labs in general (perhaps helped by the distillation efforts Anthropic detailed above) squeezing margins just as users are growing more price-sensitive. The frontier still sits pretty clearly in the U.S., but the volume play is still very much up for grabs.

QUICK HITS

▸ Rich built a streaming recommender that knows who's on the couch

Today’s workflow comes from reader Rich Kroll : “I built Show Hole to solve my family's classic streaming problem: we spend too much time deciding what to watch, and the answer changes depending on who is actually on the couch.

We each have different tastes, but the more interesting problem is that our tastes overlap differently in different combinations. I watch one kind of thing with my spouse, my spouse watches something different with our kid, and my kid and I have our own lane too. Most recommendation tools flatten that into one account profile, one watch history, or broad genre buckets, so they miss the real context of a household.

See Rich’s full workflow Visit The Rundown University . How do you use AI? Tell us for a chance to be featured.

🐳 DeepSeek V4.1 Flash - DeepSeek's small, efficient new open model

⚙️ SWE-2 - Cognition’s powerful new coding model inside of Devin

🗣️ GPT-Live-1 - OpenAI's voice model that listens while talking, now on API

🎨 Krea Agents - Creative agent with memory, style skills, Slack/Figma hookups

Cognition rolled out SWE-2 inside Devin, a Kimi K3-based coding model that claims to match Fable 5.1 and GPT-6 Astra on certain coding benchmarks while costing 64% less.

Thinking Machines co-founder Andrew Tulloch is moving from Meta to Anthropic, months after initially rejecting Mark Zuckerberg's reported $1.5B package and then signing on anyway at an undisclosed compensation.

OpenAI denied to the NYT that Tristan Buckmaster's Codex prompts shaped its Navier-Stokes proof, while claiming "substantial progress" on another Millennium Prize problem.

OpenAI launched ChatGPT for Financial Services, an edition of ChatGPT Work with PitchBook, Crunchbase, and LSEG data baked in for valuation models and pitch decks.

Universal Music Group is partnering with AI audio startup ElevenLabs on a licensing deal, with a fan platform for remixing participating artists’ tracks in development.