Forcepoint X-Labs has released a new post by researcher Jyotika Singh, which is a deep dive into the accelerating AI Cybersecurity Arms Race. The post details how artificial intelligence is simultaneously empowering defenders with real-time detection while helping adversaries automate deception at massive scale. The central finding is that every algorithm built for protection can now be turned to exploit, making speed and continuous adaptation the only sustainable advantages. This analysis highlights that the challenge for security leaders is no longer whether to use AI, but how to stay ahead of sophisticated, AI-enabled adversaries.
Key highlights from the research include:
- Adversaries are leveraging malicious LLM variants (such as FraudGPT and WormGPT) to automate phishing kit creation, malware generation, and massively scale social engineering operations.
- Deepfake technology has fully graduated from theory to multi-million-dollar real-world fraud, exemplified by a confirmed £20M video-call scam that impersonated company officers for a fraudulent transfer.
- Attackers are using Reinforcement Learning (RL) to train generative models to automatically evolve polymorphic payloads, creating malware that changes structure to evade endpoint security products.
- Defenders are fighting back with multi-layered ML/DNN classifiers and ‘Agentic AI’ systems, cutting average dwell time by automating real-time threat detection and high-volume tasks like alert triage.
- Actionable recommendations for organizations, including enforcing out-of-band verification for high-value transfers and continuously red-teaming internal ML models against adversarial inputs.
This research reinforces that the future of cybersecurity will be decided by who adapts the fastest, and that human oversight paired with intelligent automation is critical to maintaining confidence in protection.
The full blog post can be found at https://www.forcepoint.com/blog/x-labs/ai-cybersecurity-arms-race
NATO’s Biggest Naval Exercise Proves Undetectable Ship-to-Ship Laser Communication
Posted in Commentary with tags Astrolight on October 17, 2025 by itnerdLithuanian space and defense tech company Astrolight has successfully demonstrated undetectable, unjammable, and high-bandwidth laser-based ship-to-ship communication with its POLARIS terminal during REPMUS’25, NATO’s largest unmanned maritime exercise recently.
During the REPMUS (Robotic Experimentation and Prototyping using Maritime Uncrewed Systems)/Dynamic Messenger mission, hosted by the Portuguese Navy, POLARIS laser terminals maintained a stable, jam-proof horizon-limited laser-based link between two vessels: NRP Dom Francisco de Almeida and NRP Dom Carlos I. During testing, the link wasn’t detected by a single sensor of other participating ships, drones, and land assets.
Astrolight’s terminals also transmitted gigabytes of data at latencies and speeds that allow for more than 10 concurrent, real-time HD video streams, even through rain and fog, during the day and night.
Jamming is a serious problem at sea because it can distort satellite navigation, confuse radar and ship-tracking displays, and interrupt radio and satellite communications. In such cases, crews switch to less secure backup methods like noisy radio or signal lamps that increase a ship’s electromagnetic signature and make it easier to detect.
The demonstration of Astrolight’s POLARIS in Portugal builds on prior tests with the Lithuanian Navy.
NATO’s REPMUS/Dynamic Messenger exercise combines REPMUS, the top event for maritime robotics and unmanned tech, and Dynamic Messenger, a program for testing innovative naval systems. They bring together NATO Allies, partners, academia, and industry experts, and provide a realistic setting to evaluate new maritime capabilities and promote their integration into NATO operations.
Leave a comment »