A sophisticated cyber-espionage toolkit can disappear for decades-and still outperform modern implants.In this Briefing, we will present the discovery and reconstruction of an undocumented nation-state framework believed to have gone extinct morethan 20 years ago. Although buried in time, this "fossil" malware reveals engineering decisions that remain surprisingly competitivetoday, and it appears to have quietly influenced reconnaissance platforms that emerged long after it.But the story isn't only about the tool. Because this framework was never publicly documented, it creates a rare testbed for somethingsecurity teams are increasingly relying on: AI-driven malware analysis and attribution. We use the case to measure how modern AI systems behave when they cannot rely on training data, prior writeups, or recognizable patterns-and compare their conclusions to careful, expert-led reverse engineering. The result is both a recovered chapter of cyber-espionage history and a grounded look at where AI helps, where it misleads, and what its failure modes look like in cyber-paleontology.