Perplexity AI: Search Revolution or Compliance Risk?
Let me share with you my short article on the subject of Perplexity. I hope you will find it worth reading. I would highly appreciate your feedback. The short poll is at the end of the post. ✔️Is your team already using Perplexity AI? Perplexity (note: this isn't a language model) could change how we work with information, but it also brings serious risks. ✔️Scale of Usage 780 million queries per month with 20% month-over-month growth. Perplexity's current valuation hovers around $18 billion, and since investors are "voting with their dollars," they clearly see this as the future of search tools. It's a US-registered company with mostly Venture Capital funding (60%). 22 million monthly active users. The question is whether companies are consciously supporting this usage or if it's "shadow AI" as described in the interesting MIT report "The GenAI Divide: State of AI in Business 2025"? ✔️Why is Everyone Talking About This? For managers, it means competitive analysis in minutes instead of hours, meeting prep with real-time market data, and competitive intelligence with source citations. For IT teams, it delivers 1,200 tokens/second (3x faster than competitors). What's especially important is that one API gives you access to 8+ models (GPT-5, Claude 4.0, Gemini 2.5), for just $5 per million tokens vs. $30 at OpenAI!!! ✔️Architecture Their proprietary Sonar model is built on Llama 3.3 70B, specialized for factual responses and real-time internet search. Thanks to Cerebras infrastructure, it achieves record-breaking speeds of 1,200 tokens per second. They use DeepSeek R1 - Chinese reasoning technology, but running locally on Perplexity servers in the US. According to sources, data "never" leaves American infrastructure, eliminating the risk of transfer to China. An OpenAI-compatible API ensures easy integration with existing systems, eliminating the need for code changes.