OpenAI Goes Enterprise

OpenAI is pushing deeper into corporate tech with a new Infosys alliance, a privacy tool for scrubbing data, and a reported financing vehicle. (infosys.com) Infosys said on April 22 that it will combine OpenAI models and products including Codex with Infosys Topaz Fabric, its agentic artificial intelligence services suite. The companies said the early focus is software engineering, legacy modernization, DevOps automation, and e-commerce. (infosys.com) OpenAI also released Privacy Filter on April 22 under an Apache 2.0 license on GitHub and Hugging Face. OpenAI said the model runs locally, so personal data can be masked or redacted on a laptop or in a browser before that text is sent anywhere else. (openai.com) Privacy Filter is a smaller, single-purpose model built to find names, addresses, and other personal identifiers in long documents. OpenAI’s GitHub page says it has 1.5 billion total parameters, 50 million active parameters, and a 128,000-token context window. (github.com) The common problem here is deployment, not demos. Big companies have spent the past year testing coding assistants and chatbots, but many still need systems integrators, internal controls, and privacy tooling before they can use them across regulated workflows. (techcrunch.com) That helps explain a third piece of the story: capital. Mint, citing the Financial Times, reported that OpenAI is in talks to back a $10 billion vehicle called DeployCo, with an initial $500 million stake and the option to add another $1 billion later. (livemint.com) Reuters, citing the Financial Times, reported that TPG, Bain Capital, Advent International, Brookfield, and Goanna Capital are among the outside investors expected to put in about $4 billion. Reuters said it could not independently verify the report and that OpenAI and the firms did not immediately respond to requests for comment. (finance.yahoo.com) The reported structure goes beyond selling model access through an application programming interface, or API, which is the standard software pipe companies use to buy AI. It suggests OpenAI is also testing a model that bundles software, implementation partners, and financing for enterprise rollouts. (techcrunch.com, livemint.com) Taken together, the moves point to the same target: making OpenAI easier to buy, safer to use, and harder to rip out once it is inside a company’s core systems. The next test is whether large customers move from pilot projects to signed, scaled deployments. (infosys.com, openai.com, livemint.com)