Nathalie Trepanier spent years as a reporter, covering crime for a number of media outlets in Ottawa and elsewhere. As a journalist, what she saw online bothered her.
Already a Subscriber? Log in
Get Instant Access to This Article
Become an Ottawa Business Journal Insider and get immediate access to all of our Insider-only content and much more.
Nathalie Trepanier spent years as a reporter, covering crime for a number of media outlets in Ottawa and elsewhere. As a journalist, what she saw online bothered her.
“I kept seeing lies being spread across social media with impunity,” she told OBJ. “I mean, it was just striking to me because as a journalist that doesn’t compute.”
That frustration gave rise to SPEAC Social, an Ottawa-built platform that allows users to run social posts through an AI fact-check. The beta is live at speacsocial.com in nine countries, and Trepanier is now raising a $750,000 seed round.
The name SPEAC is an acronym Trepanier worked backwards to create. “I kept thinking knowledge is power, knowledge is power,” she said. “That term was coined in the 1700s by a philosopher who spoke Latin, ‘scientia potentia est.’
“And then I started seeing where this was going and it turns out that ‘accuracy is key’ in Latin is ‘accuratio clavis,’ or ac. So knowledge is power, accuracy is key,” she continued. “So there you have SPEAC.”
Here’s how SPEAC Social works. A user writes a post. Before it goes anywhere, SPEAC’s AI engine checks it against a hierarchy of sources weighted by editorial standards. International scientific consensus and peer-reviewed material sit at the top, while news sources with high trust and strong correction records sit below that. Wikipedia is excluded. The system returns three sources for each finding.
Each post then gets one of five labels: verified, disputed, contested, needs review, or opinion. Verified content pushes a user’s trust score up and gets posted immediately. Repeated disputed posts drag the score down. Users with disputed claims are asked if they want to change their post before hitting publish, and the published post is clearly labelled as disputed once it goes live on the platform.
Contested covers cases where credible expert opinion is genuinely divided, while needs review applies to breaking news that has not yet been indexed publicly. Opinion exempts subjective expression from the verification process but is clearly labelled as an opinion post.
The system is meant to make accuracy a kind of social currency, and to make life difficult for bots from malicious actors.
The backbone of the system is powered by two AI systems working in tandem. “It’s AI-driven,” Trepanier explained. “So I’m using dual architecture. Anthropic’s Claude is my first point and then as a backup I have (Google) Gemini. But they both follow the same prompt architecture.”
The idea for SPEAC came to Trepanier over time. Her frustration with disinformation followed her out of the newsroom when she left journalism, and continued to irk her as she watched social media progress.
“I saw a few people lying consistently and not really being called out on it,” she said. “And people bought into it and it was affecting mental health and physical health and finance and everything.”
She tried reaching out to the platforms such as Meta and X to ask for accountability, offering some ideas about how the platforms could implement checks to combat disinformation and misinformation. “Obviously a no-reply came,” she said.
Trepanier admits her work as a crime reporter took a toll on her, and it’s one of the reasons she became a passionate defender of the truth.
“I was having a nightmare every single night,” she said. “Like, I would go to bed and I’d shut my eyes and I knew without fail there would be a nightmare.”
She moved into communications, deliberately avoiding media relations. She worked across the private, public and not-for-profit sectors, including a role at the Office of the Conflict of Interest and Ethics Commissioner. She points to those experiences as the reason she can claim the editorial neutrality SPEAC requires.
“I’m not affiliated with any political party. I don’t have skin in any game other than I want accuracy to prevail,” she said. “I think editorial integrity is huge right now. I think that’s one of the biggest aspects that I bring to it is the neutrality. I don’t have economic interests in any major platform or anything.”
She’s careful about how she frames SPEAC Social and says she is not competing with the Facebooks and Xs of the world. There is no engagement model at all on SPEAC Social.
“It’s turning that on its head,” she explained. “So it’s not about engagement. It’s about accuracy, not who is speaking the loudest or the most colourful or who gets you the most excited. It’s about calm. It’s giving you a moment of pause. And I think people are there now.”
She describes SPEAC as a spell-check, but for facts.
The beta’s early traction would indicate that her idea has legs. Canada has the most users, followed by the United States, with China in third.
Trepanier says the target user is anyone whose work depends on being believed, such as journalists, medical practitioners, financial institutions and universities. She says SPEAC could plug into media literacy curricula. She flags gen Z as the demographic most receptive to the platform.
The business model has four streams. There’s a free tier capped at 25 fact-checks per day, and there’s a paid subscription that allows longer-form checking on full articles. There are institutional licences for newsrooms and universities and spots for advertisers to sit beside verified posts.
Trepanier does not have a programming background and instead contracted an outside dev shop to help build the platform, but things stalled. She ended up building the beta herself using agentic AI.
Capital, however, remains the hardest challenge. “It’s a little more challenging in Ottawa than it would be in Silicon Valley,” she said. “I’m not going to pretend otherwise. As a white francophone woman, I did not fit the mould. I don’t have a network that includes Mark Zuckerberg. So I’m still facing that challenge.”
Some early reviewers told her people would reject the product because they prefer their “echo chambers,” but Trepanier rejects the premise, especially when the stakes get personal.
“When it’s your own health, when it’s your parents’ or kids’ health, when it’s your finances or career, I don’t think so. I think people want the truth,” she said.
“A lot of people have been unwilling accomplices in sharing lies, but they don’t realize that they’re lies,” she continued. “And when they find out, they’re so grateful. Like oh my gosh, I’m glad I wasn’t part of this.”
She said image verification is coming next, and audio and video verification will follow once the underlying technology gets cheaper and more accessible.
Trepanier sees the arc of her life leading to this. “Everything was the runway to bring me to this moment. I think I’m really well-poised to be the one to introduce this to the masses.”
