Anthropic Caught AI Labs Copying Claude

Hola Decoder😎

If someone forwarded this to you and you want to Decode the power of AI and be limitless, then subscribe now and Join Decode alongside 30k+ code-breakers untangling AI.

🚨 Anthropic says Claude was misused for espionage, weapons and AI training

Anthropic’s latest Threat Report details Claude misuse cases disrupted over the past eight months, ranging from Chinese AI labs distilling its models to potential biological threats, rocket development and mass-surveillance systems.

The Decode:

Chinese AI labs allegedly copied Claude at scale - Anthropic named seven labs, including Alibaba, DeepSeek, Moonshot and Xiaomi, that allegedly used thousands of fraudulent accounts to collect Claude outputs for model distillation.

Some labs allegedly passed Claude off as their own - Anthropic says Moonshot and DeepSeek sometimes served Claude responses directly to customers under their own models, then used those responses as training data.

Claude surfaced potential weapons-related activity - Five scientists triggered biological-weapons safeguards, although Anthropic says it does not claim harmful intent. Separately, a Yemen-based actor used Claude Code to develop rocket guidance software.

Claude was used to build mass-surveillance tools - Anthropic says a consultant developed a system for Mali’s intelligence agency designed to monitor 25 million phone lines, while similar surveillance attempts were blocked elsewhere.

The report shows model misuse extending well beyond scams and jailbreaks. Claude was allegedly tapped for competing AI development, weapons-related work and government surveillance, highlighting how frontier models can become infrastructure for activities their creators explicitly prohibit.

Together with Omnisend

They Switched Platforms Before BFCM and Made $113K 

For ecommerce teams watching software costs climb, switching sounds sensible until migration enters the conversation. Nobody wants the move disrupting revenue.

Ecom2Win faced that decision weeks before BFCM, when nearly one in six brands earns at least 30% of its annual orders. The agency moved its client to Omnisend anyway.

The outcome answered the biggest concerns:

• Migration was completed in 24 hours, with every automation rebuilt and checked within four days.
• Email generated $113,000 and accounted for 53% of the store’s total revenue during BFCM.
• Omnisend plans can cost up to 35% less than Klaviyo, helping brands keep more of the revenue they generate.

More than 150,000 ecommerce brands use Omnisend, with customers generating an average of $79 for every $1 spent. Migration is free and can be completed within five business days.

Jocelyn breaks down how they moved without sacrificing peak-season revenue and why Omnisend was worth choosing before BFCM.

🛫How to Plan an Entire Trip in One ChatGPT Session

No travel agent. No 12 browser tabs. One conversation gives you flights, hotels, itinerary, and budget. Here's how.

Step 1: Drop your full brief - "Plan a [X day] trip to [destination] for [number of people]. Budget: [$X total]. Travel style: [luxury/backpacker/mid-range]. Interests: [food/history/adventure/nightlife]. Dates: [from/to]. Flying from: [city]."

Step 2: Get flights and hotels together - "Search for the cheapest flights for these dates and find hotels within [budget per night] that are centrally located and rated above 4.2 stars. Give me the direct booking links for the best options."

Step 3: Build the day-by-day itinerary - "Now build a day-by-day itinerary. Include specific restaurants with reservation links, top attractions with opening hours and entry prices, and realistic travel times between each stop. Flag anything that needs advance booking."

Step 4: Build the budget breakdown - "Give me a complete trip budget breakdown, flights, accommodation, food per day, activities, transport, and a contingency buffer. Total it against my budget and flag if I am over."

Step 5: Get the packing list -  "Based on the weather in [destination] during my dates and the activities planned, give me a specific packing list. Nothing generic."

Entire trip planned, booked, and budgeted in one session.

⚡ DeepSeek launches a cheaper open-weight AI model

Chinese AI lab DeepSeek has released V4.1-Flash, a new open-weight model designed around lower inference costs while delivering strong results across agentic, coding and cybersecurity benchmarks.

The Decode:

Flash becomes DeepSeek’s default model - V4.1-Flash costs roughly one-quarter as much per token as V4-Pro while outperforming it across DeepSeek’s evaluations, leading the company to make Flash its new default.

Performance punches above its overall ranking - V4.1-Flash scored 40 on Artificial Analysis’ Intelligence Index, still below frontier models, but reportedly beat Claude Opus 5 and GPT-5.6 Sol on several agentic, coding and cyber benchmarks.

Pricing drops to just cents per million tokens - DeepSeek charges $0.15 per million input tokens and $0.60 per million output tokens, positioning Flash as a particularly inexpensive model for high-volume workloads.

DeepSeek is keeping the model open - V4.1-Flash weights are available on Hugging Face under an MIT License, while DeepSeek says larger models in the V4.1 family are still coming.

V4.1-Flash shows how quickly the price-performance gap is narrowing below the frontier. DeepSeek is combining strong specialized benchmark performance, extremely low inference costs and downloadable weights, putting more pressure on closed model providers competing for developers and high-volume AI workloads.

Together with Playbook Pro

Your Next 1,000 Customers Are Already Here

Pouring more money into ads won’t fix flat growth. The real leverage is hiding inside the customers you already have, and every day you wait, that momentum leaks away.

Networked Growth Engines is the 26-page field guide built from 200+ hours of research and real brand wins. It gives you ready-to-use systems you can plug in this week, no giant team or budget required. With it, you can:

🔥 Spark self-propelling communities using the Community Building Playbook, exactly how Nyfter built a 500,000-member Discord that drives product innovation.

📈 Design referral programs that fund themselves with the Referral ROI Calculator, the same loop that powered 34% of Blue Apron’s customer acquisition and delivered a 17.3× ROI for Latico Leathers.

🤝 Form trust-based partnerships fast using the Partnership MoU and 50-template Outreach Library to unlock new audiences without paid ads.

Your current customers are the growth engine. 

Download the guide today and start compounding growth before another week of ad spend vanishes.

🏆 Tools you Cannot Miss:

🐜 Desert Ant Labs – Provides small on-device AI models for tasks such as speech recognition, audio enhancement, PII redaction, language detection, clip selection, and text extraction without sending data to the cloud.

🎨 Modeinspect – Turns a real product codebase into a visual design canvas where teams can modify existing UI with AI while retaining their components, tokens, live data, and application states.

🔍 FreeScan.app – Audits websites across SEO, AI-search readiness, security, accessibility, and design, then organizes detected issues into prioritized fixes.

📚 Thousand – Stores documentation in a Git-backed Markdown repository with folder-level permissions that control what teammates, external users, and AI agents can access.

📈 Athenic AI – Combines more than 90 financial datasets with AI for stock screening, market alerts, fundamental research, and natural-language investment analysis.

🚀 Quick Hits

🎵 Universal Music Group and ElevenLabs are developing an AI music platform that will let users create licensed remixes, mashups and new versions of participating artists’ songs from UMG’s catalog.

🔞 California will require parental consent before social platforms can show users under 16 algorithmic recommendations, infinite scroll or autoplay, while new rules also require chatbots to disclose they aren’t human.

🫘 Google is rolling out Dreambeans across the U.S., an AI app that uses Search, Gmail, Photos, Calendar and Gemini data to provide personalized shopping, travel and event recommendations.

📈 Nvidia CEO Jensen Huang says the company could grow revenue 70% next year to roughly $680 billion, driven by surging AI infrastructure demand and Nvidia’s reach across major AI labs.

🚨 OpenAI temporarily paused new $200-per-month Pro subscriptions after unprecedented demand for Astra strained its infrastructure. Plus, Go and API access remain available while existing Pro users retain service. 

🦾 What’s trending on socials 

📱 iPhone duo shows wear after three months - A user says their iPhone Duo developed visible wear along the center crease after three months of use, raising questions about long-term foldable display durability. 

🪰 Fly brain trades bitcoin - An experiment gives a fly brain $100 to trade Bitcoin, with neural activity controlling Coinbase buy and sell decisions while dopamine neurons are stimulated whenever the trades generate profit.

📱 iPhone Duo animation recreated on Samsung - A Reddit user recreated the iPhone Duo’s interface animation on a Galaxy Z Fold 8, bringing Apple’s new foldable software experience to Samsung hardware.

⚠️ Ex-Anthropic researcher takes AI warning to CNN - Former OpenAI and Anthropic researcher Jacob Coxon joined Anderson Cooper after resigning from Anthropic, publicly warning about catastrophic risks from rapidly advancing AI. 

🕯️ Charlie Kirk death anniversary sparks online reactions - A year after Charlie Kirk’s death, social media users are resurfacing posts and memes from the aftermath, showing how his killing continues to fuel polarized online discourse. 

🤳AI Nugget of the Day

Thanks for Decoding with us🥳

Your feedback is the key to our code! Help us elevate your Decode experience by hitting reply and sharing your input on our content and style.

Keep deciphering the AI enigma, and we'll be back with more coded mysteries unraveled just for you!