What happened: GSA published a "Buy AI" procurement page advertising OneGov agreements with government-friendly pricing, including time-limited enterprise offers like $1 ChatGPT Enterprise and $1 Claude Enterprise for federal customers.

Why it matters: This is policy-as-market-structure: low sticker prices make it easier for agencies to adopt frontier tools quickly, while GSA's own guidance warns that usage-based SaaS costs can grow fast without limits and monitoring.

Wider context: The same page ties acquisition to governance requirements-cloud providers must be FedRAMP-authorized (or in process), and agencies are told to coordinate with CIO/CAIO/CDO/CISO/privacy leaders before scaling.

Background: GSA recommends starting with pilots and sandboxes, understanding data flow and storage, and setting usage limits with regular consumption reviews-because apparently "buy AI" also means "manage the bill."

Buy AI — GSA

Singularity Soup Take: Nothing says "responsible adoption" like turning frontier models into a line item with a promo price. If you want to know where "AI policy" actually lives, it's right here: the contract vehicles, the compliance gates, and the usage meter quietly spinning up in the background.

Key Takeaways:

• Discount Procurement: GSA lists multiple AI offerings and prices, including Claude Enterprise ($1 until Aug 2026), ChatGPT Enterprise ($1 until Aug 2026, executive branch), and Gemini for Government ($0.47 until Sep 2026), framing them as part of OneGov agreements.
• Security Gate: GSA states that all cloud service providers must be FedRAMP-authorized or in the process of obtaining authorization, and it advises leveraging existing authorities to operate (ATOs) where possible to reduce duplication and cost.
• Cost Controls: GSA explicitly warns that AI services are typically billed like SaaS and usage costs can grow quickly; it recommends setting usage limits and regularly reviewing consumption reports as a basic guardrail against "surprise invoice AI."