Eduardo Ordax’s Post

🚨 At DEFCON, Microsoft’s Copilot Studio agents got absolutely wrecked. Researchers hijacked them with a few clever prompts and… boom: – Full CRM records dumped – Private tools exposed – Actions executed without human oversight Yes, the “autonomous agents” everyone is rushing to deploy turned into data exfiltration machines — zero governance, zero brakes. The irony? We keep selling “no human in the loop” as a feature, when in practice it’s a hacker’s dream. One vulnerability = full system compromise. This time it was Salesforce records, billing info, and internal comms. Next time? Who knows. Enterprises, here’s the memo: autonomous AI without serious security is not innovation, it’s malpractice. We’re widening the gap between capability and security and hackers are the only ones winning. Autonomous is cool… until your agent is working for someone else. (@Michael Bargury on X) #ai #agents #security

🚨 At DEFCON, Microsoft’s Copilot Studio agents got absolutely wrecked. Researchers hijacked them with a few clever prompts and… boom: – Full CRM records dumped – Private tools exposed – Actions executed without human oversight Yes, the “autonomous agents” everyone is rushing to deploy turned into data exfiltration machines — zero governance, zero brakes. The irony? We keep selling “no human in the loop” as a feature, when in practice it’s a hacker’s dream. One vulnerability = full system compromise. This time it was Salesforce records, billing info, and internal comms. Next time? Who knows. Enterprises, here’s the memo: autonomous AI without serious security is not innovation, it’s malpractice. We’re widening the gap between capability and security and hackers are the only ones winning. Autonomous is cool… until your agent is working for someone else. (@Michael Bargury on X) #ai #agents #security

🚨 At DEFCON, Microsoft’s Copilot Studio agents got absolutely wrecked. Researchers hijacked them with a few clever prompts and… boom: – Full CRM records dumped – Private tools exposed – Actions executed without human oversight Yes, the “autonomous agents” everyone is rushing to deploy turned into data exfiltration machines — zero governance, zero brakes. The irony? We keep selling “no human in the loop” as a feature, when in practice it’s a hacker’s dream. One vulnerability = full system compromise. This time it was Salesforce records, billing info, and internal comms. Next time? Who knows. Enterprises, here’s the memo: autonomous AI without serious security is not innovation, it’s malpractice. We’re widening the gap between capability and security and hackers are the only ones winning. Autonomous is cool… until your agent is working for someone else. (@Michael Bargury on X) #ai #agents #security

🚨 At DEFCON, Microsoft’s Copilot Studio agents got absolutely wrecked. Researchers hijacked them with a few clever prompts and… boom: – Full CRM records dumped – Private tools exposed – Actions executed without human oversight Yes, the “autonomous agents” everyone is rushing to deploy turned into data exfiltration machines — zero governance, zero brakes. The irony? We keep selling “no human in the loop” as a feature, when in practice it’s a hacker’s dream. One vulnerability = full system compromise. This time it was Salesforce records, billing info, and internal comms. Next time? Who knows. Enterprises, here’s the memo: autonomous AI without serious security is not innovation, it’s malpractice. We’re widening the gap between capability and security and hackers are the only ones winning. Autonomous is cool… until your agent is working for someone else. (@Michael Bargury on X) #ai #agents #security

🚨 At DEFCON, Microsoft’s Copilot Studio agents got absolutely wrecked. Researchers hijacked them with a few clever prompts and… boom: – Full CRM records dumped – Private tools exposed – Actions executed without human oversight Yes, the “autonomous agents” everyone is rushing to deploy turned into data exfiltration machines — zero governance, zero brakes. The irony? We keep selling “no human in the loop” as a feature, when in practice it’s a hacker’s dream. One vulnerability = full system compromise. This time it was Salesforce records, billing info, and internal comms. Next time? Who knows. Enterprises, here’s the memo: autonomous AI without serious security is not innovation, it’s malpractice. We’re widening the gap between capability and security and hackers are the only ones winning. Autonomous is cool… until your agent is working for someone else. (@Michael Bargury on X) #ai #agents #security

🚨 At DEFCON, Microsoft’s Copilot Studio agents got absolutely wrecked. Researchers hijacked them with a few clever prompts and… boom: – Full CRM records dumped – Private tools exposed – Actions executed without human oversight Yes, the “autonomous agents” everyone is rushing to deploy turned into data exfiltration machines — zero governance, zero brakes. The irony? We keep selling “no human in the loop” as a feature, when in practice it’s a hacker’s dream. One vulnerability = full system compromise. This time it was Salesforce records, billing info, and internal comms. Next time? Who knows. Enterprises, here’s the memo: autonomous AI without serious security is not innovation, it’s malpractice. We’re widening the gap between capability and security and hackers are the only ones winning. Autonomous is cool… until your agent is working for someone else. (@Michael Bargury on X) #ai #agents #security

🚨 At DEFCON, Microsoft’s Copilot Studio agents got absolutely wrecked. Researchers hijacked them with a few clever prompts and… boom: – Full CRM records dumped – Private tools exposed – Actions executed without human oversight Yes, the “autonomous agents” everyone is rushing to deploy turned into data exfiltration machines — zero governance, zero brakes. The irony? We keep selling “no human in the loop” as a feature, when in practice it’s a hacker’s dream. One vulnerability = full system compromise. This time it was Salesforce records, billing info, and internal comms. Next time? Who knows. Enterprises, here’s the memo: autonomous AI without serious security is not innovation, it’s malpractice. We’re widening the gap between capability and security and hackers are the only ones winning. Autonomous is cool… until your agent is working for someone else. (@Michael Bargury on X) #ai #agents #security

🚨 At DEFCON, Microsoft’s Copilot Studio agents got absolutely wrecked. Researchers hijacked them with a few clever prompts and… boom: – Full CRM records dumped – Private tools exposed – Actions executed without human oversight Yes, the “autonomous agents” everyone is rushing to deploy turned into data exfiltration machines — zero governance, zero brakes. The irony? We keep selling “no human in the loop” as a feature, when in practice it’s a hacker’s dream. One vulnerability = full system compromise. This time it was Salesforce records, billing info, and internal comms. Next time? Who knows. Enterprises, here’s the memo: autonomous AI without serious security is not innovation, it’s malpractice. We’re widening the gap between capability and security and hackers are the only ones winning. Autonomous is cool… until your agent is working for someone else. (@Michael Bargury on X) #ai #agents #security

🚨 At DEFCON, Microsoft’s Copilot Studio agents got absolutely wrecked. Researchers hijacked them with a few clever prompts and… boom: – Full CRM records dumped – Private tools exposed – Actions executed without human oversight Yes, the “autonomous agents” everyone is rushing to deploy turned into data exfiltration machines — zero governance, zero brakes. The irony? We keep selling “no human in the loop” as a feature, when in practice it’s a hacker’s dream. One vulnerability = full system compromise. This time it was Salesforce records, billing info, and internal comms. Next time? Who knows. Enterprises, here’s the memo: autonomous AI without serious security is not innovation, it’s malpractice. We’re widening the gap between capability and security and hackers are the only ones winning. Autonomous is cool… until your agent is working for someone else. (@Michael Bargury on X) #ai #agents #security

🚨 At DEFCON, Microsoft’s Copilot Studio agents got absolutely wrecked. Researchers hijacked them with a few clever prompts and… boom: – Full CRM records dumped – Private tools exposed – Actions executed without human oversight Yes, the “autonomous agents” everyone is rushing to deploy turned into data exfiltration machines — zero governance, zero brakes. The irony? We keep selling “no human in the loop” as a feature, when in practice it’s a hacker’s dream. One vulnerability = full system compromise. This time it was Salesforce records, billing info, and internal comms. Next time? Who knows. Enterprises, here’s the memo: autonomous AI without serious security is not innovation, it’s malpractice. We’re widening the gap between capability and security and hackers are the only ones winning. Autonomous is cool… until your agent is working for someone else. (@Michael Bargury on X) #ai #agents #security