Wire · regulatory
GSA Issues a Revised Proposed Clause Governing AI Data-Safeguarding Requirements
◆ Sectors
◆ Geography
◆ Source
◆ Verified
Fusion42 · 20 July 2026 · Fusion42 review
The GSA has issued a revised proposed contract clause (552.239-7001) that narrows the scope of AI data-safeguarding requirements to apply only when government data is processed through large language models, but expands its reach across all GSA-administered contracts rather than just MAS contracts. The clause removes the government's unlimited "any lawful purpose" license and certain content-moderation restrictions from the earlier March draft, with comments due 3 August 2026.
This Wire brief sits within Fusion42's coverage of AI Frontier Models and Cybersecurity.
◆ ◆ The Wire takeaway
If you sell anything to the US government and use an LLM to process or deliver work, you now have a compliance obligation that lands on any contract with GSA, not just AI deals - and you have until 3 August to comment before this becomes binding. The good news: the government gave up unlimited access to your AI system, but the bad news is the rule now applies much wider than the original draft, catching service providers who thought they were out of scope.
◆ Coverage
1 source · 20 Jul 2026
◆ Related on Wire
◆ Topics