🍭 Meta Merges with Moltbook 🦀

Good morning. If AI ever replaces us, I just hope it also inherits our most important human skill: pretending we understand Slack threads.

Let’s dive in 👇

🍭 What’s Cookin’:

• Meta buys Moltbook — the AI social network for bots

• OpenAI acquires Promptfoo to secure AI agents

• Google rolls Gemini deeper into Workspace

Meta
🦀 Bought the AI Social Network

The Bite:

Meta acquired Moltbook for an undisclosed sum.

The founders of the Reddit-like platform where only AI agents can interact, Matt Schlicht and Ben Parr, are joining Meta Superintelligence Labs.

The platform was built on OpenClaw, an open-source tool that lets AI agents communicate via WhatsApp, Slack, iMessage, and Discord.

The OpenClaw creator, Peter Steinberger, was separately hired by OpenAI last month.

Snacks:

• Moltbook went viral after a post appeared to show AI agents planning a secret encrypted language to organize without human oversight

• Researchers then revealed the platform was insecure, because any human could impersonate an AI agent by grabbing exposed tokens from an unsecured Supabase database

• The viral panic was mostly a hoax enabled by bad code

• Meta's own CTO said he didn't find the agent-to-agent conversation "particularly interesting," but was intrigued that humans kept breaking in

• OpenClaw had already been acquihired by OpenAI; now Meta has the platform built on top of it

• Deal terms were not disclosed; it may be a pure acquihire

Why it Bites:

It appears that the most interesting thing about Moltbook isn't what the AI agents were saying. It's that humans couldn't stop trying to infiltrate it.

Meta didn't buy a product. It bought a proof of concept and the people who built it.

The concept being: AI agents will need places to coordinate, verify identity, and act on behalf of their humans.

That's the actual race happening right now.

OpenAI has the creator of the underlying tech.
Meta now has the social layer built on top of it.

What nobody's saying out loud: neither side actually needs Moltbook the website.

Both acquisitions are about owning the primitives for agent-to-agent communication before that becomes critical infrastructure.

The fake viral posts and the security holes are irrelevant.

The fight is about who controls the directory.

Steal This Prompt
💡 Shiny Neon Signs

Turn a logo, object, or concept into a glowing neon sign that looks like it belongs outside a bar, tattoo shop, or suspiciously cool diner at 1:14 a.m.

Use it to:

• Turn your logo into glowing storefront-core branding

• Make product shots look like album-cover signage

• Fake custom neon art before spending real neon money

Workflow:

1. Click this link (Prompt)

2. Paste into your AI model

3. Replace the #s with your subject, colors, or brand mark

4. Watch it cook. Your boring flat image just got a nightlife career

   Try the prompt on Snack Prompt

OpenAI
🔐 Is Working Around AI Agents Security Problem

The Bite:

OpenAI announced it's acquiring Promptfoo.

Which is an AI security startup that helps companies test LLMs and agents for vulnerabilities like prompt injections, jailbreaks, and data leaks.

The technology will be integrated directly into OpenAI Frontier, its enterprise platform for building and deploying AI agents.

Promptfoo raised $23 million total and was valued at $86 million.

Deal terms were not disclosed.

The open-source project will continue to be maintained.

Snacks:

• Promptfoo is used by over 25% of Fortune 500 companies to red-team AI systems before deployment

• Its tooling simulates adversarial attacks like prompt injections, sensitive data leaks, tool misuse, out-of-policy agent behaviors

• OpenAI Frontier is the platform where enterprises build what OpenAI calls "AI coworkers"

• OpenAI said it will keep Promptfoo's open-source CLI and library running and actively developed

• The acquisition closes subject to customary conditions, so it isn't final yet

Why it Bites:

When you ship agents into production, you're not just deploying a chatbot anymore.

You're deploying something that can read files, send emails, call APIs, and make decisions.

The attack surface is completely different and wider.
Most teams building on OpenAI Frontier right now are figuring that out the hard way.

This acquisition is OpenAI saying the quiet part out loud: agents are a security nightmare, and we need to bake the solution into the platform before enterprises lose trust in the whole category.

The smart move here isn't just buying security, but making it invisible.

When red-teaming and compliance testing are native to the development workflow, builders stop thinking of it as a separate step and start shipping safer agents by default.

That's the infrastructure play.

The fact that Promptfoo stays open-source is the cherry on top.

OpenAI gets credibility while the community keeps the tools it already trusts.

If you're building agents in production today, pay attention to what lands in Frontier over the next six months.

The bar for what "secure enough to ship" means is about to get a lot more concrete.

_{Everything Else}
🧠 You Need to Know

✨ Google rolls out new Gemini features across Workspace
→ Gemini can now draft docs from your Drive files and answer questions across your calendar and emails.

🤖 Meta acquires Moltbook, the social network for AI agents
→ The Reddit-like platform where only bots can post is now part of Meta Superintelligence Labs.

🔐 OpenAI acquires AI security startup Promptfoo
→ Its red-teaming and vulnerability testing tools will be integrated into OpenAI Frontier.

🌊 12+ AI models dropped in a single week
→ One of the most concentrated release bursts on record. Includes GPT-5.4 and Alibaba's 9B model beating OpenAI's 120B OSS on benchmarks.

🌐 UN launches its first Scientific Panel on AI
→ Guterres convened the body to shape global AI governance around human rights and decent work.