On August 19, 2026, AgentMail experienced an outbound email disruption lasting approximately eight hours, from 6:40 a.m. to 2:35 p.m. PDT. During this period, customers were unable to send email through the affected service.
We are sorry for the disruption and the impact it had on our customers.
What happened
A small number of accounts used AgentMail to send unsolicited email, violating our acceptable-use policies. Although we had systems in place to detect and suspend abusive accounts, these senders distributed their activity across multiple accounts and avoided our existing account-level limits.
Amazon Web Services (AWS), our email infrastructure provider, monitors and moderates traffic sent through its platform. This activity damaged the reputation of our shared email infrastructure, leading AWS to temporarily pause outbound sending for the entire service. This affected legitimate customers as well as the accounts responsible for the abuse.
Why the disruption lasted nearly eight hours
Restoring service required AWS to review our corrective measures and reinstate our sending access. Due to circumstances on AWS's side, we did not have access to the escalation channels that would normally help us resolve a critical issue promptly. This extended the time required to reach the appropriate review team and restore service.
How we resolved it
We identified and suspended the accounts responsible, blocked the abusive sending patterns, and worked with AWS to restore service. Once the case reached the appropriate review team, AWS reinstated our sending access. Outbound sending resumed at approximately 2:35 p.m. PDT.
How we are preventing a recurrence
We have strengthened our safeguards in several areas:
- New accounts have stricter limits on how many unique recipients they can contact.
- Suspicious or unwanted email content is blocked before it is sent.
- Accounts are suspended immediately when abuse, bounce, or complaint thresholds are reached.
- Unsubscribe requests, bounces, and complaints are enforced automatically to prevent further delivery.
- Sending behavior is monitored across related accounts so limits cannot be bypassed by creating multiple accounts.
- We are improving tenant-level isolation so an abusive sender can be stopped without affecting other customers.
We will continue investing in abuse prevention, monitoring, and infrastructure isolation. AgentMail is intended for legitimate, consent-based email, and protecting that ecosystem is essential to the reliability of our service.
