Pinned post
One company, two brands. MPCH is the cryptographic authority behind institutional tokenization and AI execution. Ti22, an MPCH company, brings human control and privacy to defense, intelligence, and institutions.
Ti22
@Ti22Social
Ti22, an MPCH company. Human control and privacy over money, agents, and data for defense, intelligence, and institutions.
101 Following 10 Followers 10 Posts
Latest posts 5
AI agents can isolate devices, block users and change system settings. They shouldn't decide their own authority.
We're working with Socura to test ATLAS in real security operations, enforcing human authorization before an agent acts.
prnewswire.com/news-releases/…
We're working with Socura to test ATLAS in real security operations, enforcing human authorization before an agent acts.
prnewswire.com/news-releases/…
www.prnewswire.com MPCH and Socura Join Forces to Keep AI Agents Under Human Control /PRNewswire/ -- MPCH, the top provider of high-assurance cryptographic security infrastructure, today announced that it will join forces with Socura, a... New research: 17 fixed advisories in 3 agent products. In 12, the approval didn't cover the change the agent made. In 5, the facts changed after approval.
Approve once, act later is broken. Bind each approval to the exact action and check it at write time.
#AIAgents #AISecurity
Approve once, act later is broken. Bind each approval to the exact action and check it at write time.
#AIAgents #AISecurity
.@CrowdStrike just shipped an AI Kill Switch that terminates unauthorized agents in real time.
A kill switch is the last line of defense. The first is a provable yes before each consequential action, recorded outside the agent.
#AIAgents #AgentSecurity
A kill switch is the last line of defense. The first is a provable yes before each consequential action, recorded outside the agent.
#AIAgents #AgentSecurity
Goldman just pulled Claude Code from traders using it near client flows and positions. Right call, wrong fix. A coding agent runs with whatever access its user has. Don't ban agents. Scope them: approved actions, recorded outside the agent. #AIAgents #AISecurity