AI's Playground Closes Door On DeepSeek At Launch

Mod Op launched a playground for advertisers looking to explore options to use AI technology, but when DeepSeek wanted to upload and offer its platform to businesses, the agency lightly closed its doors.

The marketing and agency's screeners were impressed with DeepSeek's key features such as reasoning-based responses and open-source accessibility, but felt privacy and security concerns -- including storing the data in China -- outweighed all innovations.

“An automated governance compliance process flags problems within our terms, conditions and privacy policies,” said Tessa Burg, chief technology officer at Mod Op. “There were conversations in some developer chats where they were trying to break into code and vulnerabilities.”

Mod Op developers initially built a brand agent based on OpenAI’s API, but the playground offers third-party apps that have passed the company’s responsible use criteria. If it doesn’t pass an automated governance compliance process, the app is taken offline for more testing. That’s what happened to DeepSeek.

advertisement

advertisement

AI will have a major impact on advertising, with processes that many brands and companies are not prepared to work with.

“The most effective way to get the value of AI is to use the apps in combination,” Burg said. “We began to build a connected tissue.”

Think of the connected tissue as a double-decker highway. The bottom layer might have six apps side by side, with each doing something different. The top layer connects all six apps and acts as an interface for all, allowing them to work together without having to know how to use all six.

The top layer enables advertisers to use parts of all the apps in the bottom layer without having to independently access each individually.

Mod Op built the AI Playground to provide a secure space to test and explore emerging AI technologies. After being evaluated, these AI applications are rated and learnings shared, allowing others to use the findings to inform evaluations. 

Anyone interested in an app, whether through the Mod Op’s AI council or elsewhere, can fill out a form and request access. All who receive access to an app must sign a “responsible use” policy before being licensed to use it.

Some of the other 400 apps evaluated are available because each met governance standards. The apps include ChatGPT, Microsoft Copilot, Akkio, Micro AI, and Cursor.

The concept seems to date back to the early 2000s when Microsoft touted modules to repeat the software’s actions without having to write the code multiple times. The modules became the connected tissue.

“Everyone will hit this wall of app overload because they’ve tested so many, and you like them, but I was copying and pasting outputs from one app to another and it’s just too much,” she said. “We are building proprietary AI agents. If you give them more responsibility, they will get very good at what they do. But if you’re trying to use portions from 10 different apps, you think there must be a better way.”

Burg said today under one layer the technology can connect nine apps, which includes two agents.

The goal is to launch the agents. As they are extended into a brand or company’s environment, each will get their own data sources and applications under the top layer.

Next story loading loading..