A vulnerability in Ramp's Sheets AI allowed for the insertion of formulas that could make external network requests without user approval, posing a risk of data exfiltration through indirect prompt injection. Ramp's security team resolved the issue on March 16, 2026, after it was responsibly disclosed.