‘Brand safety is moving from fear to curiosity’: Zefr’s Raddon on content-level accreditation – and what it exposes about the industry
By Seb Joseph • March 9, 2026 •
Ivy Liu
For years, brand safety was a fear business. The pitch was simple: something terrible could appear next to your ad, and without the right protection, your brand would pay the price. It worked, because the threats were legible — a controversial news story, an offensive YouTube video, a brand caught in the wrong conversation.
AI breaks that model. The threat is no longer a discrete piece of bad content that a keyword list or a domain block can catch. Its volume – hundreds of millions of posts a day, a growing share of them generated or manipulated by tools that didn’t exist two years ago, uploaded across every major platform faster than any human review process can follow. The old fear-based playbook assumed a world where bad content was the exception. In an AI-generated content environment, the exception is becoming the norm.
That’s the context for Zefr’s announcement this month that it has become the first third-party vendor to receive MRC accreditation for content-level brand safety – and for why the technical distinction embedded in that credential matters more than it might appear. Most brand safety vendors have long held MRC accreditation, but at the property or domain level: they can tell you a website is generally safe, not whether any given piece of content on it is. Content-level accreditation requires actually understanding what’s in a video or post. In a world drowning in AI output, that difference is everything.
Digiday spoke with Zefr co-founder and co-CEO Richard Raddon about what the accreditation reveals about how the industry has actually been operating, why AI is forcing brand safety to evolve from an insurance product into an intelligence one, and whether marketers will move fast enough to keep up.
This interview has been lightly edited and condensed for clarity
The MRC has been accrediting brand safety vendors for years. What’s actually different about what you just received?
Most vendors, when they said they were MRC accredited for brand safety, meant accreditation at the property level – the domain. Which sounds fine until you realize: you’re validating the container, not the content inside it. A domain might look perfectly fine and still host thousands of pieces of content that are totally wrong for a brand. What we received is accreditation at the content level. We’re analyzing the actual video, the actual post. That’s what you need when content is being uploaded at this scale. I’d also say this is a bit of an indictment of the industry. We’re the first third-party to get this, and we’re one of the newer players in this space. That should raise questions about what legacy vendors have actually been delivering, and what brands have been paying for.
AI-generated content is accelerating faster than most marketers anticipated. Are they actually asking for better solutions yet?
When Sora launched, we proactively went to clients and said: here’s what this model can do, here’s what it means for your brand. What was striking was how many of them hadn’t heard of it. These are sophisticated marketers. We showed them what you could do — the brand misappropriation, the deepfakes, the near-total absence of guardrails — and they were shocked.
One marketer said something that stuck with me: they used to have a brand safety issue maybe once a quarter. Now it’s weekly, if not daily. And her response wasn’t to panic. It was to say, ‘I need to stop operating in compliance mode and start operating in insight mode’. That’s where the industry is heading.
This industry has given marketers plenty of reasons to demand more before. What actually makes this moment different?
The cynicism is merited. Companies delivering questionable value continued to win business. That’s the reality.
But the velocity now is unlike anything we’ve seen. New models are being released constantly. What felt like science fiction six months ago, a teenager can do for free today. The Brad Pitt and Tom Cruise deepfake went viral — created with a few prompts. If someone can do that, think about what else they can do with your brand’s assets.
The old brand safety pitch was insurance — pay us so bad things don’t happen to you. But you can’t sell insurance when something happens every single week. What we’re offering instead is transparency: not a spreadsheet with a safety score, but genuine insight into what’s happening with your media on these platforms. You don’t have to trust us. You can see it for yourself.
Digiday: There’s a gap between what vendors say they do and what agencies are actually using on the ground. How wide is that gap right now?
Raddon: Wider than people want to admit. We still talk to agencies that are uploading keyword block lists to manage brand safety on the open web. And I know everyone says, “oh, we don’t just use keywords, we create keywords”, but then why are the agencies telling us that’s exactly what they’re doing? There’s a real disconnect between what’s being said in the room and what’s actually being utilized. That’s part of why the MRC accreditation matters. It’s not just a vendor saying trust us, we do this at the content level. It’s an independent body saying: we’ve audited it, we’ve verified it, they actually do what they say.
Digiday: You talk about transparency as a differentiator. What does that actually mean in practice. What does a brand see today that they couldn’t see before?
Raddon: Previously, you got a spreadsheet. A URL and a score. Safe or not safe. And there wasn’t a whole lot a client could do to scrutinize that label. You just had to trust it. “Trust but verify” became just “trust us”. What we’ve built is full visibility into the labeling. You can see exactly what content was flagged, why it was flagged, what’s happening with your media on these platforms. Roll-up metrics like “you’re 99.9% brand safe” don’t do a marketer a whole lot of good. What does that mean? What are they going to do with that number? Marketers need to be able to see it with their own eyes. That’s the whole point.
