A billion AI tokens passed through TrustedRouter in a single day, yet the Miami startup said it could not read the prompts moving through its system.
That mix of scale and privacy is central to founder Joseph Perla’s case for a new layer of AI infrastructure. Instead of sending every request to one model, TrustedRouter directs each task to the option that best fits its cost, speed, and security needs.
The company has raised a $1.25 million seed round from investors including Sam Lessin of Slow Ventures, early Zoom and Canva investor Bill Tai, 23andMe co-founder Linda Avey, and former Stripe data leader George Xing.
Headquartered in Brickell, TrustedRouter gives developers access to more than 600 models from over 81 providers through one application programming interface. Its routing code is open source, allowing technical teams to review how the system works.
The product addresses a problem facing companies that want employees to use AI without giving every task to a large and costly model.
Businesses were “struggling with the double-edged sword of how do I get my employee base up to speed so I don’t lose, but how do I not blow my annual budget in six weeks?” Tai shared in an interview with Axios.
Within that setting, the need for routing was “universal,” he said.
TrustedRouter treats AI models as options for different types of work. A powerful model might handle a hard coding request, while a smaller, cheaper model could process a basic classification task. Developers can switch providers without rebuilding their applications each time.
“AI applications should not be locked to one model. Different jobs need different models, and switching should not require rebuilding the application or giving up control of private data,” Perla wrote in a LinkedIn announcement.
✓
Check your inbox for a confirmation email.
Perla is a four-time exited founder and a former employee of Facebook and Lyft.
Access to many models creates its own security issue. A routing service can become another place where confidential prompts, customer information, and internal records are exposed.
Joseph Perla
TrustedRouter said its open-source system uses confidential computing so the router cannot view customer prompts. Businesses can also choose which AI providers may receive their information for certain tasks.
Lessin called routing an important part of the AI story because it gave customers a “trusted, private, secure way” to use AI tools without leaking data.
Interest in this market has grown fast. Stripe recently agreed to acquire OpenRouter for more than $8 billion, Axios reported, while Meta was said to be developing a competing service called Switchboard.
Routing could also change the balance of power among AI companies. If software chooses a model for each request based on price or performance, providers may have to compete for each task instead of relying on customers tied to one system.
“I see people excited about us and moving to us because they don’t like what Anthropic is becoming,” Perla said.
TrustedRouter has been in public beta since May. Its billion-token day showed early demand, but the company must now turn that activity into steady growth while proving its privacy claims at a larger scale.
“This is day one,” Perla asserted. “We have a lot to prove, and we’re ready to do the work.”
READ MORE IN REFRESH MIAMI:
I am a Miami-based technology researcher and writer with a passion for sharing stories about the South Florida tech ecosystem. I particularly enjoy learning about GovTech startups, cutting-edge applications of artificial intelligence, and innovators that leverage technology to transform society for the better. Always open for pitches via Twitter @rileywk or www.RileyKaminer.com.
Latest posts by Riley Kaminer (see all)