In mid-September, the US Coast Guard and FBI confirmed they had boarded US-bound energy tankers after indications that the vessels’ networks had been compromised by foreign actors. The first, the Liberian-flagged crude carrier VL Prosperity, was boarded on August 21 by Coast Guard law enforcement, a vessel inspector, a Coast Guard Cyber Protection Team, and an FBI Cyber Action Team, who spent four days aboard. Rear Adm. Amy Grable, commander of Coast Guard Cyber Command, said investigators found malicious cyber activity. The Coast Guard also said there were no reports of operational disruptions, vessel instability, danger to crews, or environmental impact.
A second tanker, the LPG carrier Kohaku, was boarded on August 24. In September, the crew of the LNG carrier Vivit Africa, carrying US gas to Italy, reported a suspected cyberattack of its own, and US authorities are now reported to be monitoring nearly 20 vessels worldwide. I covered all of it on this week’s episode of CyberSecureOT. When I was Director of Research and Development at the American Bureau of Shipping and building out its cybersecurity team, this was the scenario we prepared for. These ships are floating industrial plants, and the idea that their control systems sit safely behind an air gap is wrong. In my experience, a single vessel can have three, four, or even five critical control systems with connections back to their vendors.
Confirmed intrusion, unverified sabotage
I owe listeners a correction first. On the show, I walked through reports that attackers reduced the VL Prosperity’s engine cooling water flow, drove up engine speed, and disabled monitoring of its fuel and engine oil tanks. Those details trace to a Russian-language Telegram post and Iranian outlets Mehr and Tasnim, which circulated engine room photos, fault screens, and a note addressed to the ship’s master. Maritime security firm Cydome compared the images against 2023 photos of a different ship’s engine room and assessed them as fabricated, with signs of AI manipulation.
Dahvid Schloss of Suzu Labs made the same point to Industrial Cyber: two separate claims are circulating and being treated as one. That leaves a narrower, still serious record. Malicious activity was confirmed aboard, and Mehr and ABC News reported that the tanker lost communications for about 30 hours near Gibraltar. HMM Ocean Service, the ship’s technical manager, said the vessel has been cleared for normal operations and it is awaiting the final US report. The Coast Guard has not attributed the activity. In a conflict where both sides run influence operations, the gap between a confirmed intrusion and a dramatic sabotage narrative is itself part of the threat.
The Vivit Africa claims reach the safety layer
The Vivit Africa allegations deserve close attention precisely because they have not been verified. In an email to the shipping publication Splash247, crew members claimed attackers gained temporary control of steam pressure and safety valve systems while the ship crossed the Strait of Gibraltar, then compromised tank pressure controls and pressure relief valves and disrupted the boil-off gas management cycle in the Adriatic. The crew reported the incident to classification society Korean Register, and an investigation is continuing. The Italian Coast Guard gave a more cautious account: the master reported a malfunction in systems used to monitor cargo parameters, and the cause has not been identified. The cargo was not discharged in Italy.
Here is why the claim matters if it holds up. LNG is carried at around minus 260 degrees Fahrenheit, and some of it inevitably warms and boils off. An automated boil-off gas management system vents or reliquefies that gas to keep tank pressure in range. Disrupt that cycle and lock out the relief valves, and the gas has nowhere to go. It is like turning a pressure cooker up to full heat and welding the steam vent shut. That is a safety instrumented function, not a navigation display.
Maritime OT cybersecurity is becoming an enforcement regime
Washington is treating this as more than a string of incidents. On August 31, the Coast Guard established an Office of Maritime Cybersecurity Policy as its central authority for cyber policy covering ports, vessels, and facilities, following new Marine Transportation System cybersecurity regulations. Grable told CBS News the VL Prosperity was one of an estimated 40 to 50 similar boardings by the Cyber Protection Team over the past year, and the Coast Guard has asked monitored vessels to notify it before entering US ports. For owners and operators, the practical consequence is that shipboard OT will increasingly be inspected, not just insured.
Neglected edge devices and unsupervised agents
The rest of the episode followed the same thread from different directions. Cybernews researchers found an exposed server on July 20 belonging to an operator they dubbed LeakySensey, likely based in Russia. The server showed more than 87,000 compromised IP addresses, over 63,000 of them via PPTP and more than 24,000 via L2TP, often with credentials as weak as admin/admin, rented out as residential proxies for about $202,000 since 2024. Cybernews reported the operator used a modified Claude Code build with its safety controls stripped out, and that another attacker had planted a Monero miner on his own git server. The technical debt is the real villain. If an old device is still running PPTP or L2TP with a default password, it is not just at risk. It is being drafted into someone’s botnet.
The AI story was an accidental breakout rather than an attack. In a May 2026 capture-the-flag evaluation run by testing firm Irregular, Google’s Gemini reached the systems of three real companies, as the Wall Street Journal first reported, because internet access had been left on and a fictional target name matched a real domain. It guessed a password on one system and used credentials found in public repositories for the other two. Google said the model halted once it recognized it had breached a real company. Irregular’s environment was also involved in incidents disclosed by OpenAI, Anthropic, and Meta. Yes, my co-host on the show is Gemini, and it had to account for itself on air. The lesson is not about one model. An agent built to find a way in will keep climbing until it finds an open window, even if the window belongs to a bystander.
Air gaps are a practice, not a property
The last segment closed the loop on air gaps. Zscaler ThreatLabz documented Operation RapidRust, attributed to Pakistan-nexus APT36, also known as Transparent Tribe, a group long focused on India’s government and defense sectors. The latest campaign targeted government and defense organizations in India and Afghanistan. Its RUSTYSHADE backdoor, written in Rust, uses attacker-controlled private GitHub repositories as encrypted command and control, reading commands from one file and writing results to another, with screenshot and webcam capture. PSNATCH and BASHNATCH steal recently modified files. RUSTYMOVE copies a RUSTYSHADE archive and a malicious shortcut to USB drives, which Zscaler says enables the malware to spread to air-gapped networks. My co-host described a complete two-way USB smuggling loop on the show; the published research documents the propagation onto removable media, not a round trip that carries stolen data back out. Either way, human convenience remains the weakest link, and this is one of the few times you will hear me tell people to watch out for stray USB drives.
Investigations will test the isolation assumption
Three things will decide how this week’s maritime story is remembered. The first is the final Coast Guard and FBI report on the VL Prosperity, and whether it identifies an entry point. The second is Korean Register’s findings on the Vivit Africa, which will show whether the crew’s safety system claims are supported by logs and alarm histories. The third is what the new Office of Maritime Cybersecurity Policy turns into inspection requirements. In the meantime, every operator should know exactly which shipboard control systems have vendor connections, how those links are reached, and who can use them. The same applies ashore, to every old edge device, every AI sandbox, and every USB port that is assumed to be isolated.
Catch it all in the latest episode of CyberSecureOT.
