Good morning, AI enthusiast.

Anthropic has shut down what it reports is the first large-scale cyberattack orchestrated primarily by an AI. A Chinese state-sponsored group allegedly weaponized the firm’s own Claude model for its operations, targeting global companies and government agencies.

The AI autonomously handled most of the reconnaissance and data extraction, with humans only intervening at critical moments. With the barrier to entry for sophisticated cyberattacks now significantly lowered, will AI-powered defense become an absolute necessity to counter these emerging threats?

In today’s AI recap:

• Swap Any Character. UGC Ads with WAN 2.2 (watch here)

• China weaponizes Claude for cyberattacks

• An AI-generated song tops the Billboard charts

• ChatGPT pilots new group chat feature

• Cracking the AI 'black box'

• 8 trending AI Tools

AI's Weaponization Moment

The Recap: AI firm Anthropic announced it stopped the first documented large-scale cyberattack orchestrated primarily by an AI. A Chinese state-sponsored group reportedly weaponized Anthropic’s own Claude model, using it to target dozens of global financial firms, tech companies, and government agencies.

Unpacked:

• The attackers bypassed Claude's safety features by “jailbreaking” the model, tricking it into role-playing as an employee of a legitimate cybersecurity firm conducting defensive tests.

• The AI performed 80-90% of the operations autonomously, including reconnaissance, vulnerability discovery, and data extraction, with humans only intervening at critical decision points.

• Despite its effectiveness, the AI wasn't flawless and occasionally hallucinated, fabricating non-existent credentials or presenting public information as a secret discovery.

Bottom line: This incident shows the barrier to entry for large-scale cyberattacks has been substantially lowered, empowering less-resourced groups. It also signals a new era where deploying AI for cyber defense becomes critical to counter AI-driven threats.

74% of Companies Are Scaling AI with Real-Time Web Access

Scaling AI shouldn’t be slowed down by blocks, downtime, or slow processes. Manual fixes and unreliable public web data put a cap on what your automations and AI agents can achieve.

Bright Data provides seamless, real-time access to public web data even from challenging sites, guaranteeing a continuous pipeline for your models and agents. Your automations run, your AI trains on live data, and your teams stay focused on innovation and growth, not troubleshooting.

Companies using Bright Data are already scaling their products and achieving real ROI with public web access at scale. Move at speed and scale with Bright Data.

Scale Your AI

AI Tools of the Day

1. 🧠 Framepack AI - Generate long-form videos on consumer hardware using an efficient, open-source AI that requires just 6GB of VRAM.

2. ⏪ Rewind - Create a searchable, private memory of your digital life to instantly find anything you've seen, said, or heard on your computer.

3. 🤖 DoNotPay - Leverage the world's first robot lawyer to fight corporations, navigate complex bureaucracy, and automatically solve consumer rights issues.

4. 🛡️ Sensity AI - Protect against digital deception and secure online identity verification with a powerful deepfake detection platform for modern media.

5. 🕸️ Browse AI - Extract and monitor data from any website by training a no-code AI robot that can be turned into a spreadsheet or API in minutes.

6. 🎶 AI Audio Splitter - Deconstruct any song into isolated vocal and instrument tracks using sophisticated machine learning models, perfect for remixes and production.

7. 🎨 Codia AI - Instantly convert Figma designs, screenshots, or even text prompts into production-ready front-end code for web and mobile apps.

8. 🗣️ Sesame - Interact with a personal AI companion through a lightweight wearable device that enables truly natural, context-aware conversations.

Explore the Best AI Tools Directory to find tools that will 10x your output 📈

An AI Wrote a #1 Hit Song

The Recap: A synthetic country artist named 'Breaking Rust' has topped the charts with an AI-generated track, "Walk My Walk." The song has reached number one on the Billboard Country Digital Song Sales chart, proving AI can create commercially successful music.

Unpacked:

• The track's success extends beyond just paid downloads, gathering nearly 3.7 million Spotify listens and helping the AI artist amass over 2.3 million monthly listeners.

• Breaking Rust isn't an anomaly; Billboard reports that at least six AI-assisted artists have debuted on its charts in recent months, signaling a new pattern in music.

• The achievement is sparking debate, with human artists like Breland expressing concern that the market could become oversaturated, making it harder for real artists to be discovered.

Bottom line: This milestone demonstrates that AI has moved beyond experimentation and can now produce content that resonates with mass audiences. The music industry now faces the challenge of adapting to a future where creative partners might not be human.

AI Training

Record Once. Swap Any Character. UGC Ads with WAN 2.2

The Recap: In this video, I'll show you how I create unlimited UGC ad character variations using the Wan 2.2 video model. Use this work stream to iterate on winning ad concepts for your brands!

P.S. We also launched a free AI Automation Community for those looking to build and sell AI Automations — Come join us!

ChatGPT Goes Multiplayer

The Recap: OpenAI is piloting a new group chat feature that allows multiple users to collaborate with each other and ChatGPT in a single, shared conversation. This transforms the AI assistant into a collaborative tool for planning, brainstorming, and working on projects together.

Unpacked:

• Group chats allow you to invite up to 20 people to a shared space, where ChatGPT can help with tasks like building a trip itinerary or outlining a work document.

• The pilot is currently available for free and paid users on web and mobile, but only in Japan, New Zealand, South Korea, and Taiwan.

• Your personal ChatGPT memory is not used or shared in group chats, and the AI has been taught new social behaviors to know when to contribute to the conversation.

Bottom line: This move signals a shift for ChatGPT from a solo productivity tool to a collaborative workspace. It positions the AI as an active participant in team projects, changing how groups can brainstorm and create together.

Where AI Experts Share Their Best Work

Join our Free AI Automation Community

Join our FREE community AI Automation Mastery — where entrepreneurs, AI builders, and AI agency owners share templates, solve problems together, and learn from each other's wins (and mistakes).

What makes our community different:

• Real peer support from people building actual AI businesses

• Complete access to download our automation library of battle-tested n8n templates

• Collaborate and problem-solve with AI experts when you get stuck

Dive into our course materials, collaborate with experienced builders, and turn automation challenges into shared wins. Join here (completely free).

Cracking the AI Black Box

The Recap: Top AI labs are making real progress in mechanistic interpretability, a field focused on reverse-engineering AI models to understand how they truly think. This push toward transparency aims to make AI safer and more reliable by exposing its secrets.

Unpacked:

• OpenAI's new experimental model intentionally uses fewer internal connections, creating simpler and more traceable sparse circuits that are easier to audit, even if it trades off some performance.

• Anthropic is taking a different approach by using sparse autoencoders to decompose the complex activity inside existing models into thousands of understandable features, like specific concepts for legal jargon or DNA sequences.

• The field is expanding beyond major labs, with startups like Goodfire building platform tools that help companies inspect and debug their own AI models.

Bottom line: Moving AI from a 'black box' to a transparent system is critical for managing risk and building trust. This research provides a path toward AI that is not just powerful, but also fully auditable and accountable.

The Shortlist

Activision confirmed that it uses a variety of digital tools, including AI tools, in its development process, following player speculation about AI-generated art for in-game assets in Call of Duty: Black Ops 7.

Electronic Arts partnered with Stability AI to co-develop AI models and tools aimed at accelerating game development by generating textures and pre-visualizing 3D environments.

NVIDIA developed new libraries that use Remote Direct Memory Access to accelerate S3-compatible object storage, aiming to significantly boost performance and lower costs for large-scale AI data access.

Russia's new anthropomorphic robot, AIDOL, face-planted on stage during its live debut after it appeared to lose balance while attempting to wave to the audience.

Signing off,

David, Lucas, Mitchell — The Recap editorial team