Hardening the standard by which AI agents talk to each other — public funding aimed at an open-source specification itself
The National Science Foundation awarded $1.5 million to the University of Illinois at Urbana-Champaign to improve the safety, security and trustworthiness of the Natural Language Interaction Protocol, an emerging open standard that lets AI agents exchange structured messages. It falls under Safe-OSE, a program aimed at open-source ecosystems themselves.
Grant overview (primary data)
- Award amount$1,500,000
- RecipientUniversity of Illinois at Urbana-Champaign (Illinois)
- ProgramSafeOSE
- Period2026-09-01 〜 2028-08-31
- FunderU.S. National Science Foundation (NSF) / NSF
Key points
- An award to improve the safety, security and trustworthiness of NLIP, an emerging open standard for AI agents to exchange structured messages ($1,500,000; 2026-09-01 to 2028-08-31).
- The institution is the University of Illinois at Urbana-Champaign, under Safety for Open-Source Ecosystems (Safe-OSE).
- AI agents are being deployed across government and business systems but lack a safe way to communicate across organizations and technologies.
- The design hardens the safety side of a standard already in use rather than building something new.
- Across the 120 NSF awards this site holds as of 2026-09-02, the median obligated amount is $1,955,351 and the median duration three years; this award is modest on both.
1More agents means a protocol is needed
AI agents are appearing across government services, business systems and research environments. As their number grows, what becomes the problem is not the intelligence of any single one but how they interact across organizational and technological boundaries.
Without a shared procedure by which agents built separately by different organizations can safely exchange structured messages, each either stays isolated or accumulates ad hoc connections. NLIP, the subject of this award, is an emerging open standard attempting to define that shared procedure.
2Funding aimed at the safety side of a standard
Safe-OSE stands for Safety for Open-Source Ecosystems, a framework weighted toward making already widely adopted public infrastructure safe rather than building something new. Weaknesses in open standards are often found after adoption has spread, by which point users are numerous. Working on the reliability of the communication layer while adoption is beginning addresses that ordering problem directly.
3The place of the communication layer
Agent safety tends to bring to mind inspecting model outputs. But where several agents work together, the properties of the communication layer decide overall safety: who a message came from, whether its content was altered in transit, whether it reached an unintended party. Lining up individually safe agents does not guarantee safety in combination if the exchanges between them are not verified.
4Scale and duration
The obligated amount is $1,500,000 over two years. Across the 120 NSF awards this site holds as of 2026-09-02, obligated amounts have a median of $1,955,351 and durations a median of three years, so this award is modest on both counts. There are 70 distinct programs at that date and Safe-OSE is one. Given the subject, outcomes will appear not as new capability but as an existing ecosystem that breaks less easily. Amounts are the obligated amount as of the check date and may change.
Why it matters
Where agents are combined to run work, overall safety depends on the communication layer between them as much as on each model. Public money going to the safety side of an open standard is also a case in point for how foundational software is sustained.
FAQ
What is NLIP?
What is Safe-OSE?
Sources (primary)
Source: NSF Award Search (U.S. National Science Foundation, public domain). Amounts are the obligated amount. For privacy, we do not handle principal investigator names.
- NSF Award (original, official)
- NSF Award ID: 2625398