Good morning, AI enthusiasts. Mozilla just deployed its entire $1.4B reserve balance to fund an AI safety "rebel alliance" while the viral assistant Clawdbot racks up 44,200 GitHub stars despite security experts warning it could hijack your computer through a single DM.
With big tech burning billions on compute and open-source tools raising major security questions, is Mozilla's bet on safety governance too little, too late?
In today's recap:
Mozilla's $1.4B all-in bet on AI safety startups
Viral Clawdbot assistant raises serious security concerns
Turn yourself into a 3D model
Albania's AI minister undermined by developer corruption
4 new AI tools and more
MOZILLA
Mozilla deploys $1.4B to challenge OpenAI, Anthropic
Source: Reuters
Recaply: Mozilla just deployed roughly $1.4B in reserves to fund what president Mark Surman calls a "rebel alliance" of startups focused on AI safety, transparency and governance, positioning the Firefox maker as a counterweight to dominant AI labs.
Key details:
Mozilla Ventures launched in 2022 with $35M and has invested in 55+ companies to date, with the organization now tapping its full reserve balance to expand the initiative focused on open-source AI ecosystems.
OpenAI and Anthropic have raised over $60B and $30B respectively, commanding valuations of $500B and $350B, while Mozilla holds $1.4B in total reserves from two decades as a nonprofit browser company.
Surman runs the operation from a farm outside Toronto and told CNBC he's playing the long game, with the goal of funding a "mainstream" open-source AI ecosystem for developers by 2028.
The Trump administration has criticized AI safety efforts as "woke AI" and signed an executive order establishing a task force to challenge state AI regulations, creating headwinds for Mozilla's governance-focused strategy.
Why it matters: Mozilla's betting its entire reserve balance on AI at a time when Firefox continues losing ground to Chrome and Edge. The $1.4B figure sounds impressive until you compare it to the $90B+ that OpenAI and Anthropic have collectively raised, revealing a David-versus-Goliath dynamic where Mozilla's fighting not just for AI safety but for relevance itself. Whether a 20-year-old browser nonprofit can meaningfully shape the AI landscape against labs burning billions annually on compute remains an open question.
OPEN SOURCE & ANTHROPIC
Viral AI assistant Moltbot raises security red flags
Source: Moltbot
Recaply: Moltbot, the open-source personal AI assistant formerly called Clawdbot until Anthropic forced a rebrand, has amassed over 44,200 GitHub stars in weeks, but security experts warn its utility-versus-risk trade-off could turn ugly for users approaching it like ChatGPT.
Key details:
Austrian developer Peter Steinberger built the tool to "manage his digital life" and explore human-AI collaboration, with users able to prompt it through WhatsApp, Telegram, Signal, Discord and iMessage to execute tasks like calendar management, email sending and terminal commands.
Cloudflare stock surged 14% in premarket trading on January 27 as social media buzz around the AI agent resparked investor enthusiasm for Cloudflare's infrastructure, which developers use to run Moltbot locally on devices.
CEO of SocialProof Security Rachel Tobac warned that giving Moltbot admin-level access plus app credentials creates major risks, noting that "if I can interact with it by DMing you on social media, well now I can attempt to hijack your computer in a simple direct message."
Steinberger himself messed up the renaming process, with crypto scammers snatching his GitHub username to create fake cryptocurrency projects, prompting him to warn followers on January 27 that any project listing him as coin owner is a scam.
Why it matters: Moltbot's captured developer imagination by promising the "AI that actually does things" vision that Siri launched in 2011 but never delivered. The hype's real: thousands of stars on GitHub, stock moves for Cloudflare, viral social posts showing off setups. But there's a security-versus-utility paradox: running it safely means isolating it on a separate computer with throwaway accounts, which defeats the purpose of having a useful AI assistant. For now, it's a power tool for experienced developers willing to manage the risks, not the mainstream personal assistant the hype suggests.
GOVERNMENT AI
Albania's AI anticorruption minister undercut by graft accusations
Source: AP
Recaply: Albania created the world's first AI-generated government minister named Diella to eliminate bribery in document processing, but some of the developers behind the anticorruption avatar have now been placed under house arrest for suspected corruption themselves.
Key details:
Diella, named after the Albanian word for sun, was designed to assess government contract bids and help eliminate bribery when citizens apply for government services online through Albania's digital portals.
The developers placed under house arrest for corruption accusations are part of the humans-in-the-loop team that built and maintains the AI model, raising questions about whether the minister can be programmed to overlook certain evidence of graft.
Albania positioned Diella as a public anticorruption crusader and a safeguard against corrupt practices, with the AI model intended to provide transparency in government operations that have historically struggled with graft issues.
The timing of the corruption accusations against developers calls into question the achievability of the AI model's core mission, since the same people accused of corrupt practices were responsible for programming the anticorruption safeguards.
Why it matters: There's a certain irony in building an AI to fight corruption when the builders themselves are accused of the very thing they're trying to eliminate. Albania's experiment was ambitious: using AI to create accountability where human systems had failed. But it highlights a fundamental problem with AI governance: the model's only as trustworthy as the data it's fed and the humans programming it. If the developers can influence what Diella sees or ignores, the whole anticorruption premise collapses. It's a reminder that AI can't solve institutional problems that humans haven't figured out first.
NEWS
Beyond the big stories in AI today
OpenAI introduced Prism, a free AI-native workspace for scientists powered by GPT-5.2, integrating drafting, revision, collaboration and LaTeX compilation with unlimited projects and collaborators for ChatGPT personal accounts.
Yahoo launched Scout, an AI answer engine built on Anthropic's Claude and Yahoo's 30 years of search data, with inline citations and vibrant emoji-filled results across 250 million monthly U.S. users.
Twin Labs introduced Twin, an AI company builder that deployed 100,000+ fully autonomous agents during its 1-month beta, with users building entire businesses like trading systems and plumbing operations with no code.
Tongyi Lab released Z-Image, a 6B-parameter image generation model family, with Z-Image-Turbo achieving sub-second inference on H800 GPUs and ranking as the #1 open-source model on Artificial Analysis leaderboard.
Google DeepMind launched Agentic Vision in Gemini 3 Flash, enabling code execution for image manipulation and delivering a consistent 5-10% quality boost across vision benchmarks through iterative zoom, inspect and observe loops.
LobeHub introduced agent teammates, a next-generation harness with 10,000+ skills and MCP-compatible plugins, building on its 70,000-star GitHub community with unified intelligence across models and white-box personal memory.
Google added monthly cloud credits to AI Pro and Ultra plans, providing $10 per month for Pro subscribers and $100 per month for Ultra subscribers through integrated Google Developer Program benefits.
Mistral AI launched Vibe 2.0, a terminal-based coding agent offering multi-file orchestration, smart references and async agents, with native IDE extensions for VS Code and JetBrains supporting tab completions and context-aware chat.
SoftBank is in talks to invest up to $30 billion more in OpenAI as part of a funding round that could raise $100 billion and value the company at $830 billion.
GUIDES
Turn yourself into a 3D model with Meshy AI
Recaply: In this tutorial, you will learn how to turn your selfies into detailed 3D models using Meshy AI's multi-view feature. The AI analyzes photos from different angles to create accurate models you can animate, 3D print, or use in games.
Step-by-step:
Take 2-4 photos of yourself from different angles with good lighting and a plain background. Capture front, side, and back views. Make sure each photo clearly shows your face and upper body.
Go to Meshy AI workspace and select Image to 3D. Upload your main photo. Toggle on the multi-view option to enable multiple image uploads.
Add up to 3 additional photos from different angles. Name your model something descriptive like "Jason Avatar". Click Generate to start processing.
Wait 30-120 seconds for Meshy AI to generate 4 preview models. Review each preview and select the one that best captures your likeness.
Click Generate again for final processing. Download your high-quality 3D model in GLB, FBX, OBJ, or USDZ format depending on your use case.
Pro tip: If you don't have photos from multiple angles, use ChatGPT or Meshy's built-in image generator to create reference images from different perspectives before converting to 3D.
