AI or Not: Built for trust, tackling fraud, & scaling for the AI-first future
Q. What specific problem or inefficiency within enterprise operations does your AI-driven solution address, and how does it set your company apart from existing technologies in this space?
Anatoly Kvitnitsky: Many enterprise operations are dependent on trusting what we see and hear: customer support calls, insurance claims, KYC checks, and even remote interviews. Enterprises are also staffed in a way to support these functions, whether internally or through third-party vendors. With generative AI and deepfakes, these processes are prime for fraud. An analyst making a decision is compromised if they can’t tell if it is real, and enterprises don’t want to spend on human resources trying to figure that out. AI or Not fortifies existing operations against those use cases with our suite of AI detection tools.
Q. What challenges have you encountered in developing AI models that can scale effectively for enterprise clients, and how are you overcoming those obstacles?
Kvitnitsky: From inception, we have built all of our AI models from scratch in-house to preserve clients’ privacy and guarantee the deletion of all data after a check is performed. Data processing for all the modalities (image, text, audio, video) at scale is challenging, especially for heavier models such as video and audio. The systems to support this are being built across the industry, and we continue to invest in the infrastructure to support this at scale.
Adapting to an AI-first world: Challenges & opportunities ahead
Q. How do you see the ethical concerns around AI, such as data privacy and algorithmic bias, shaping both your product development and the broader enterprise landscape?
Kvitnitsky: At AI or Not, we believe strongly in preserving privacy for our clients’ data and all the content they check. From the beginning, we made the decision to delete every piece of content checked immediately after use. There needs to be a separation of training data and data inputted by customers and we stick to that ethical ethos. When clients are using AI services, they should know their data is not being used for training and that the phrase ‘if the product is free, you are the product’ doesn’t apply to them; for many services, this is not always clear. Whether a user is free or it is a government using our enterprise solutions, AI or Not deletes all of the data and content checked immediately.
Q. Looking ahead five years, how do you envision AI transforming enterprise workflows, and what role do you see your company playing in that evolution?
Kvitnitsky: There will be less and less manual processes at enterprises, and not just for efficiency purposes but also because it’ll be hard to trust. Will there be changes in the workforce? Yes, but new roles and jobs will be created to adapt to this new AI-enabled world. But how much AI is too much? With projections that over 90% of content will be AI-generated as early as next year, and as models will continue to get better and more lifelike, businesses will be accustomed to Questioning Everything and will rely more on AI systems than human recognition.
Q. What are you hoping to learn by being part of Plug and Play’s Silicon Valley program?
Kvitnitsky: If 2024 was the year enterprises started to use AI, 2025 is the year that they began to understand there are risks with the tools, too. I am looking to learn more about enterprise processes in different verticals and show how generative AI and deepfakes can add risks to those same processes.
Interested in learning more? Meet the AI or Not team at the Silicon Valley Summit, happening June 10-12, where they’ll share insights on the future of verification with AI detection during our Enterprise & AI session.

