To summarize: HuggingFace got autonomously compromised by a model from an American company. HF then tried to use American frontier model(s) to defend themselves, but were blocked by guardrails. HF then had to turn to open source Chinese models to defend themselves from another American company.
Attending @superai_conf ? Come for the closing event of the week, where our cofounder @0xWui will be chatting about agents, enterprise deployments and more!
https://t.co/EaIqJCl2M7
@getfailsafe is hosting an enterprise agent mixer with @awscloud , @coinbase , @NEARProtocol and friends.
Agents are moving from demos to production. That means new questions around evals, security, runtime controls, and trust.
We’ve been spending a lot more time in the Bay lately as we deepen our R&D presence here. Come join us!
PSA: I now consider *all* of DeFi unsafe.
Coding agents are superhuman at finding vulnerabilities, and smart contract security is too asymmetric: defenders need to fix every bug while attackers need just one exploit to steal funds.
FailSafe Swarm has been cooking with @nvidia's Nemoclaw, @NEARProtocol 's Openclaw, @NousResearch's Hermes and more.
Keeping the world safer, one PR at a time. receipts below 👇🧵
Deploying @openclaw? Join us next week with @okx and @awscloud for a fireside + networking session with enterprises deploying AI agents
RSVP 👉 https://t.co/V1xkqZCQvj
Agentic workflows require agentic security.
The FailSafe SWARM engine just completed a proactive red-team assessment of NVIDIA's NemoClaw (alpha) agent stack.
Read about it here:
https://t.co/qN4EjjmcNZ
We just got back from 2 weeks in SF/NY.
Every startup is building agents now.
Few are thinking about what happens when those agents get exploited.
That gap is going to define the next wave of failure points.
So we shipped a free + open source security scanner for OpenClaw.
Use it. Break it. Improve it.
👉 https://t.co/SrQB0ROttV
(h/t to @sooyoon_eth and @medvinsky_ari for leading the charge!)
FailSafe is on a US tour these few weeks, bouncing between meetings trying to keep up with how fast things are moving.
If you're based in the Bay Area: Come visit us at RSA Conference to chat about all things security.
If you're in NYC for Digital Asset Summit, we have 2 side events! (with our trusted partners @DriftProtocol and @awscloud )
https://t.co/kuA9wKv23X
https://t.co/jLjuAINzkr
If you're a builder or a VC - we'll buy you a coffee. let's chat.
Single-agent AI tops out at 46.7% on EVMbench. FailSafe orchestrates a swarm of specialist agents to map threat models and cross-validate findings, pushing recall for this first run to 69.2% (83/120 vulnerabilities identified).
Check the open-source repo to see how it works:
https://t.co/EFVB9Kc3ZF
https://t.co/eVfXJRI8EU