The US Bureau of Safety and Environmental Enforcement publishes offshore incident statistics by year. Its table for calendar year 2025 shows 0 fatalities, 192 injuries, 182 fires and 2 explosions. For 2024 it shows 1 fatality, 223 injuries, 160 fires and 2 explosions. BSEE notes that counts can change after investigations and that incidents can appear in more than one category.
Fatalities are low in recent years. Fires are not: the table shows 117 in 2021 and 182 in 2025. That pattern is a useful frame for what monitoring and automation are for.
What the numbers say
| Calendar year | Fatalities | Injuries | Fires | Explosions |
|---|---|---|---|---|
| 2021 | 2 | 164 | 117 | 4 |
| 2022 | 1 | 199 | 126 | 1 |
| 2023 | 0 | 203 | 149 | 0 |
| 2024 | 1 | 223 | 160 | 2 |
| 2025 | 0 | 192 | 182 | 2 |
Source: BSEE, Offshore Incident Statistics, read 18 October 2026. The page excludes non-occupational fatalities, and some helicopter and life capsule incidents fall under other agencies.
Where remote operation and monitoring help
Offshore facilities are expensive to staff and hard to leave quickly. Software can help in a few places.
- Remote monitoring of pressure, temperature, gas and equipment health from an onshore center lets specialists watch many facilities and spot a drift early.
- Unmanned or lightly crewed facilities reduce the number of people in the hazard zone.
- Robots and remotely operated vehicles can inspect and handle tasks in areas where a person would be exposed.
- AI assisted anomaly detection can rank alarms so an operator sees the important one first.
What the software must get right
- Safe stop that works with no link. Satellite and radio links fail. Shutdown logic belongs on the platform, independent of the onshore connection.
- Human approval for changes to safety limits. A model may suggest a setpoint. A person with the right authority approves any change that touches an alarm threshold, an interlock or a shutdown.
- Secure updates. Signed firmware and software, checked against an inventory, with a rollback path. Updates reaching an offshore controller should come through a verified, controlled route.
- Tamper evident logs. Alarms, operator actions, approvals and changes should be recorded with trusted time and copied ashore, so the record survives a platform fault.
- Offline operation. The control and security stack must work with the connection down: scanning, policy checks and any AI assistance included.
- Separation between monitoring and control. A dashboard that can read data should not be able to change a setpoint.
What to do now
- Map which remote links can write to a controller and which can only read.
- Ask each vendor for an SBOM for controller and gateway software.
- Run those components against the CISA Known Exploited Vulnerabilities catalog.
- Confirm shutdown works with the satellite link removed.
- Check that logs reach shore without relying on the platform alone.
How Safeguard fits, and where it stops
Safeguard does not run a platform and this post does not claim it saves lives or reduces injuries. It offers SBOM generation and import, vulnerability scanning with known exploited status and reachability, firmware, container and dependency scanning, policy gates, audit logs, and on premises or air gapped deployment with offline models.
Try it yourself
Pick one gateway or monitoring application and ask whether a new vulnerability in a library it uses would be noticed on a schedule or only by chance. The demo below shows continuous monitoring of an inventory.
Click through it at your own pace, or open it full screen.
Quick check
Why does BSEE's table matter for a software discussion?
It shows fires rising from 117 to 182 between 2021 and 2025 while fatalities stayed low, so early detection and alarm quality remain live problems that monitoring software touches.
Where should shutdown logic live?
On the platform, independent of the satellite or radio link.
