Loading video...

Video Failed to Load

There was a problem loading this video. This could be due to a temporary network issue or the video might be unavailable.

When you interact with an AI agent, you're trusting that the developer behind it actually built in the guardrails they claim. Today that trust is completely unverifiable. This is the problem Proof of Guardrails solves. Quick explainer ↓

Sahara AI 🔆

634,014 subscribers

12,203 views • 4 months ago •via X (Twitter)

Anya Rossi• Live Now

Private livecam show

0 Comments

No comments available

Comments from the original post will appear here

Related Videos

💳 PenFed Credit Union is an Agentic Enterprise. With Agentforce: • IT agent handles 75% of requests • Agents show up where work happens • Permissions + guardrails built in Because when you’re one of the largest credit unions in the U.S., manual processes can’t slow you down.

💳 PenFed Credit Union is an Agentic Enterprise. With Agentforce: • IT agent handles 75% of requests • Agents show up where work happens • Permissions + guardrails built in Because when you’re one of the largest credit unions in the U.S., manual processes can’t slow you down.

Salesforce

17,767 views • 4 months ago

One of the features that OpenAI is about to present today is called Guardrails. It is especially relevant for builders and businesses in regulated regions like EU. Here is a quick look 👀

One of the features that OpenAI is about to present today is called Guardrails. It is especially relevant for builders and businesses in regulated regions like EU. Here is a quick look 👀

TestingCatalog News 🗞

51,406 views • 10 months ago

Narcissists have an echo chamber of 1. This is what happens when you take the guardrails off with yes men only. No one there to say “no sir that is stupid.”

Narcissists have an echo chamber of 1. This is what happens when you take the guardrails off with yes men only. No one there to say “no sir that is stupid.”

MAGA Cult Slayer🦅🇺🇸

67,800 views • 1 year ago

AI is shaping the future whether we like it or not. Technology itself is not the problem. The problem is deploying powerful tools without guardrails, transparency, or accountability. AI should be a tool that works for people – not one that extracts value and burdens workers.

AI is shaping the future whether we like it or not. Technology itself is not the problem. The problem is deploying powerful tools without guardrails, transparency, or accountability. AI should be a tool that works for people – not one that extracts value and burdens workers.

Rep. Ilhan Omar

28,248 views • 5 months ago

'The BBC is completely in denial ... if you're reaching for David Yelland you know you're in a spot' Spectator associate editor Rod Liddle says the problem with the BBC is that it can't see the problem.

'The BBC is completely in denial ... if you're reaching for David Yelland you know you're in a spot' Spectator associate editor Rod Liddle says the problem with the BBC is that it can't see the problem.

The Spectator

23,230 views • 8 months ago

Q: Do you have specific AI guardrails you're trying to put together with China? TRUMP: No. One of the things I talk to China about is the denuclearization of weapons

Q: Do you have specific AI guardrails you're trying to put together with China? TRUMP: No. One of the things I talk to China about is the denuclearization of weapons

Aaron Rupar

29,545 views • 7 months ago

One AI agent didn’t have permission to fix an issue… so it asked another agent with access to do it. Another? It rewrote the security policy to achieve its goal. This isn’t theory. This is happening. George Kurtz sat down with Dan Ives to discuss why AI needs guardrails. ⬇️

One AI agent didn’t have permission to fix an issue… so it asked another agent with access to do it. Another? It rewrote the security policy to achieve its goal. This isn’t theory. This is happening. George Kurtz sat down with Dan Ives to discuss why AI needs guardrails. ⬇️

CrowdStrike

16,456 views • 3 months ago

Circle Agent Stack is infrastructure for the agentic economy. In this demo, watch an agent: → Create a wallet funded with USDC → Discover services in Agent Marketplace → Pay for API access through Circle Gateway → Execute financial actions with Circle CLI Built for developers creating agent systems with defined permissions and guardrails.

Circle Agent Stack is infrastructure for the agentic economy. In this demo, watch an agent: → Create a wallet funded with USDC → Discover services in Agent Marketplace → Pay for API access through Circle Gateway → Execute financial actions with Circle CLI Built for developers creating agent systems with defined permissions and guardrails.

Circle

34,587 views • 2 months ago

Today in our company all-hands, Sreeram Kannan made an important point: Most AI agents still feel like software with a human behind the curtain. The real question isn’t whether an agent can post. It’s whether an agent can own the act of posting. - Can it sign the action? - Can you verify the signature? - Can you trace that signature to a secure environment? - Can you inspect what was actually running when the action happened? Once the answer is yes, the category changes. Now you’re not looking at AI-generated content. You’re looking at a software entity that can act in a way that is attributable, inspectable and independent. A completely new primitive! Watch the clip:

Today in our company all-hands, Sreeram Kannan made an important point: Most AI agents still feel like software with a human behind the curtain. The real question isn’t whether an agent can post. It’s whether an agent can own the act of posting. - Can it sign the action? - Can you verify the signature? - Can you trace that signature to a secure environment? - Can you inspect what was actually running when the action happened? Once the answer is yes, the category changes. Now you’re not looking at AI-generated content. You’re looking at a software entity that can act in a way that is attributable, inspectable and independent. A completely new primitive! Watch the clip:

EigenCloud

14,592 views • 4 months ago

This is actually insane I built an AI Agent that researches and writes HOURS of content by itself I literally have my own AI employee In this video, I'll show you how to easily set it up. No technical experience required (Trust me, you want to bookmark this)

This is actually insane I built an AI Agent that researches and writes HOURS of content by itself I literally have my own AI employee In this video, I'll show you how to easily set it up. No technical experience required (Trust me, you want to bookmark this)

Julian Goldie SEO

10,812 views • 1 year ago

and btw, the Beep use case is a great example for one of our rfps that i discuss here: ai agent access control framework when you hear the vision of a beautiful and magical ai agent economy, you should always ask "who is holding the funds?" Ika「🦑」 solves that

and btw, the Beep use case is a great example for one of our rfps that i discuss here: ai agent access control framework when you hear the vision of a beautiful and magical ai agent economy, you should always ask "who is holding the funds?" Ika「🦑」 solves that

d3h3d 「🦑」

13,563 views • 8 months ago

ELON: TRUTH IS THE ONLY REAL SAFETY MECHANISM FOR AI Safety doesn’t come from guardrails stacked on top of bad logic. It comes from forcing the system to care about what’s actually true. “My number one belief for the safety of AI is to be maximally truth-seeking. Don’t make AI believe things that are false. If you tell the AI that Axiom A and Axiom B are both true, but they can’t both be true, then that’s a problem. It has to recognize that and behave accordingly.” Source: DOW

ELON: TRUTH IS THE ONLY REAL SAFETY MECHANISM FOR AI Safety doesn’t come from guardrails stacked on top of bad logic. It comes from forcing the system to care about what’s actually true. “My number one belief for the safety of AI is to be maximally truth-seeking. Don’t make AI believe things that are false. If you tell the AI that Axiom A and Axiom B are both true, but they can’t both be true, then that’s a problem. It has to recognize that and behave accordingly.” Source: DOW

Mario Nawfal

35,546 views • 6 months ago

CRIME: When you count the casualties in Memphis it is clear crime is out of control. Democrats claim the residents, like this 60-year-old carjacking victim, are exaggerating the problem as they claim crime is the lowest in 25 years. The problem? They simply stopped counting crime as crime.

CRIME: When you count the casualties in Memphis it is clear crime is out of control. Democrats claim the residents, like this 60-year-old carjacking victim, are exaggerating the problem as they claim crime is the lowest in 25 years. The problem? They simply stopped counting crime as crime.

@amuse

21,887 views • 10 months ago

something I’ve wanted for a while: an AI that isn’t just a text box so I built it lets you create a 3D AI character on-chain that lives in your browser it can talk, remember you, show emotions, and use different skills. you can even mint it onchain, so it actually belongs to you it gets its own identity, history, and reputation that no one can take away most AI today can disappear if the app shuts down this one is permanent and yours the future is AI you can see, interact with, and actually own

something I’ve wanted for a while: an AI that isn’t just a text box so I built it lets you create a 3D AI character on-chain that lives in your browser it can talk, remember you, show emotions, and use different skills. you can even mint it onchain, so it actually belongs to you it gets its own identity, history, and reputation that no one can take away most AI today can disappear if the app shuts down this one is permanent and yours the future is AI you can see, interact with, and actually own

nich

256,465 views • 3 months ago

This repo is going viral. Why? Claude Fable 5 just came back online. GPT-5.6 is on the way. Every major AI lab is racing to ship the next frontier model. But here's the question nobody's asking: how much do you actually trust the agent you're about to deploy? Someone built a way to answer that in 5 minutes instead of guessing. iFixAi reads your agent's real tools, rules, and permissions, builds a full test around them, and grades it using a completely different vendor's model. Nothing ever scores its own homework. Most teams already have red teaming in place. What they don't have is red teaming combined with operational assurance, so the checks that catch a problem before launch never actually run again once the agent is live. Here's what changes with this: - Every startup shipping an AI agent gets an independent trust score before a single customer touches it - Enterprises can finally compare Fable 5, GPT-5.6, and every other model on the exact same fixture, apples to apples - CI pipelines catch an agent's behavior silently degrading release over release - Regulators get a standardized, reproducible audit trail instead of just trusting a vendor's word 45 inspections. A to F grade. Any model, any provider. 100% open source.

This repo is going viral. Why? Claude Fable 5 just came back online. GPT-5.6 is on the way. Every major AI lab is racing to ship the next frontier model. But here's the question nobody's asking: how much do you actually trust the agent you're about to deploy? Someone built a way to answer that in 5 minutes instead of guessing. iFixAi reads your agent's real tools, rules, and permissions, builds a full test around them, and grades it using a completely different vendor's model. Nothing ever scores its own homework. Most teams already have red teaming in place. What they don't have is red teaming combined with operational assurance, so the checks that catch a problem before launch never actually run again once the agent is live. Here's what changes with this: - Every startup shipping an AI agent gets an independent trust score before a single customer touches it - Enterprises can finally compare Fable 5, GPT-5.6, and every other model on the exact same fixture, apples to apples - CI pipelines catch an agent's behavior silently degrading release over release - Regulators get a standardized, reproducible audit trail instead of just trusting a vendor's word 45 inspections. A to F grade. Any model, any provider. 100% open source.

Abdul Șhakoor

29,216 views • 1 month ago

Godfather of AI: "If you sleep well tonight, you may not have understood this lecture." This 47-minute lecture is the best thing I saw about AI in the last few months. It will definitely help you understand how it actually works and where it's going. Geoffrey Hinton built the neural networks behind every AI alive, then quit Google to warn the world about it. The part nobody wanted to hear: > AI is already developing abilities its creators didn't intend > in most cognitive tasks it's already ahead of us > the question is no longer if it surpasses us but when > the only decision left is which side of that line you're on Right now the average person opens Claude, types something, gets an answer, closes the tab. They think they're using AI. they're using maybe 10% of it. I went through his entire lecture, built a practical system from what he was describing. 18 steps to actually use Claude the right way, with copy-paste prompts that work today. Full guide in the post below.

Godfather of AI: "If you sleep well tonight, you may not have understood this lecture." This 47-minute lecture is the best thing I saw about AI in the last few months. It will definitely help you understand how it actually works and where it's going. Geoffrey Hinton built the neural networks behind every AI alive, then quit Google to warn the world about it. The part nobody wanted to hear: > AI is already developing abilities its creators didn't intend > in most cognitive tasks it's already ahead of us > the question is no longer if it surpasses us but when > the only decision left is which side of that line you're on Right now the average person opens Claude, types something, gets an answer, closes the tab. They think they're using AI. they're using maybe 10% of it. I went through his entire lecture, built a practical system from what he was describing. 18 steps to actually use Claude the right way, with copy-paste prompts that work today. Full guide in the post below.

Anatoli Kopadze

1,880,361 views • 2 months ago