AI content memory, photo of Muslim team.
Courtesy: Tom Quisel | musubi
As policies on content review and fact-check enter a new era, a startup is turning to AI instead of humans to enforce trust and security measures.
Musubi, a startup that uses AI to adapt to online content, has raised $5 million in the seed round, the company told CNBC. The startup said the round was led by J2 Ventures, with Shakti Ventures, Mozilla Ventures and Ped seed Investor J Ventures involved.
The company was co-founded by Tom Quisel, who used to be Grinder and OkCupid. Quisel said he saw opportunities to use AI, including large language models or LLM, as well as human hosts to help social and dating apps “stay ahead” bad actors. Musubi’s AI system better understands user trends and more accurately determines bad intentions for user content.
“You’re generally hearing that trust and security teams are not satisfied with the quality of the host’s results, and you shouldn’t blame the host,” Quisel said. “It’s exactly how people just make mistakes. It’s inevitable, so, it’s true Create better job opportunities for AI and automation.”
Quisel said during Okcupid that it was a “Sisyphean fight” hosting the bad actors. Quisel said the effort required OkCupid to evacuate engineers, data scientists and other product personnel from core projects to trust and security efforts, but blocking an attack pattern never lasted long enough.
“They always figure out how to solve the defensive capabilities we have built,” he said.
Online attacks can include spam, fraud, harassment, or posting illegal or inappropriate content. This is the type of content that has been historically removed from the platform with the help of human decision makers.
Musubi claims its policy and AIMOD AI systems work together to make decisions at a 10-fold lower error rate than human moderators. The company said it also plans to use its AI to identify performance issues and inherent biases with human moderators.
Quisel said policy is the “first line of defense.” It uses LLMS to search for red flags that may violate platform policies. The red-tagged post then goes to AIMOD, which makes a modest choice to simulate what humans do with tagged posts.
Musubi’s emergence follows the industry, shifting from over-regulated online content.
The most noteworthy one is Yuan CEO Mark Zuckerberg In January Announce The company’s third-party fact-check was concluded to support a system called community annotation, which would rely on users to adjust each other’s content. This is a system first introduced by X Elon MuskWeibo service.
So far, Musubi has attracted people like Grindr and Bluesky among its customers.
The Blues need to quickly expand their adaptability after seeing it User growth soars After the 2024 election. As the Blues base grew to over 20 million users, its hosts had a lot of content to report flights. Musubi’s team of 10 people works around the clock to provide scalable solutions to the platform.
“I love that Musubi discovered the instantaneous fake and scam accounts accurately,” Aaron Rodericks, head of trust and security at Bruceki, said in a statement.
