AI or Not: Protecting Against Generative AI Risks

 

Anatoly (“Toly”) Kvitnitsky, founder and CEO of AI or Not, an AI detection startup, was interviewed on the Value Drivers podcast. Kvitnitsky's background includes a decade in fraud and KYC protection, followed by a stint as a VC at a major financial institution, which ultimately led him to start AI or Not in late 2023. His motivation stemmed from the realization that generative AI, while beneficial, could be exploited by bad actors. He felt there wasn't enough being done to protect individuals and companies from these risks and wanted to help people enjoy AI while maintaining safety and trust. Kvitnitsky notes his career has moved from M&A to startup to investment and back to startup, which gives him a unique perspective.

The interview highlighted the rapid advancements in AI and the potential dangers, especially from open source models. The emergence of DeepSeek, a Chinese open-source model that rivals the performance of models like ChatGPT at a fraction of the cost, has shown how quickly models can evolve and be adopted. Open-source models are particularly risky because they lack the safeguards of closed-source models, allowing bad actors to "jailbreak" them for malicious purposes like creating phishing emails and catfishing scams. Kvitnitsky points out that criminals are combining open-source tools to create fraud at scale, utilizing multiple modalities.

AI or Not is focused on detecting AI-generated content across different modalities, including video and audio. They provide tools for both consumers and businesses, offering a free tier for individual use and API integrations for businesses. The company believes that the problem of AI-generated content is both a B2B and consumer issue, and they are developing a multi-tiered product approach to address that. The company's focus is on identifying AI-generated content from both closed and open source models, as that is where the most risk exists. AI or Not is targeting risk and fraud groups, compliance teams, and companies that are vulnerable to AI-generated fraud, such as financial institutions, dating apps, and even medical insurance. They have even seen use cases involving fake x-rays for insurance claims .

Kvitnitsky also discussed the potential of AI to democratize content creation, allowing for high-quality content to be created at lower costs. However, he also expressed concern about the potential for the internet experience to become dominated by AI-generated content, which could create a poor experience for users . He predicted that AI-generated content will soon exceed human-generated content and emphasized the need to distinguish between real and AI-generated content. To stay ahead, AI or Not has a team of AI researchers focused on keeping up with new models and training their detection tools. They have a real-time daily checker to find new content and train their models .

AI or Not raised a $5 million seed funding round, led by Foundation Capital with participation from Plug and Play and GTM Fund . The investors knew Kvitnitsky from his previous roles, which helped the funding come together quickly. The funding is used to expand the company's efforts to keep up with the rapidly changing AI landscape. A key performance indicator for the company is usage on a per client basis, which shows whether the company is providing value.

The most challenging aspects of running his startup are hiring talent in the competitive AI space, and balancing the product roadmap with immediate customer needs.

Key Takeaways

·         Focus on a clear problem: Kvitnitsky's motivation for starting AI or Not stemmed from a clear problem, which is the potential misuse of generative AI by bad actors and the lack of sufficient protection against it. A well-defined problem helps define your company's mission and attract investors.

·         Be aware of emerging technologies and their risks: The discussion about DeepSeek highlights the importance of keeping up with rapidly evolving technologies and understanding their implications, particularly with open-source models.

.css-j9qmi7{display:-webkit-box;display:-webkit-flex;display:-ms-flexbox;display:flex;-webkit-flex-direction:row;-ms-flex-direction:row;flex-direction:row;font-weight:700;margin-bottom:1rem;margin-top:2.8rem;width:100%;-webkit-box-pack:start;-ms-flex-pack:start;-webkit-justify-content:start;justify-content:start;padding-left:5rem;}@media only screen and (max-width: 599px){.css-j9qmi7{padding-left:0;-webkit-box-pack:center;-ms-flex-pack:center;-webkit-justify-content:center;justify-content:center;}}.css-j9qmi7 svg{fill:#27292D;}.css-j9qmi7 .eagfbvw0{-webkit-align-items:center;-webkit-box-align:center;-ms-flex-align:center;align-items:center;color:#27292D;}