Britain’s intelligence apparatus has sounded an unsettling alarm: Russian propagandists are now wielding AI tools capable of digitally resurrecting the dead. The technique, which GCHQ sources describe as ‘necrobotic impersonation’, is being weaponised to sow discord by forging video and audio of deceased politicians, journalists and even private citizens. The warning, issued jointly by the National Cyber Security Centre and MI5, marks a new frontier in the Kremlin’s hybrid warfare – one that blurs the line between grief and manipulation.
At its core lies a suite of generative AI models trained on scraped social media data, voice recordings and old news footage. These systems can fabricate a realistic digital doppelgänger with chilling ease. In one documented case, a deepfake of a murdered Russian opposition figure was used to post fake confessions about funding Ukrainian saboteurs. Another involved a simulated phone call from a deceased British soldier to his mother, accusing the government of abandoning him. The psychological impact is deliberate: grieving families are weaponised, trust in recorded evidence erodes, and victims are forced to publicly disprove their own dead relatives.
The technology itself is not novel – Google and OpenAI have released similar frameworks for text and voice synthesis. But the operational twist lies in the data pipeline. Russian intelligence assets have been systematically harvesting biometric data from hacked NHS systems, genealogy forums and funeral home records. This allows them to mimic not just a face or voice, but the cadence, breathing patterns and micro-expressions that make recordings feel authentic. One intercepted file included a pixel-perfect recreation of a former MP’s speech mannerism: the way he adjusted his glasses and paused before saying ‘frankly’.
What makes this particularly insidious is the asymmetry of trust. Disinformation usually operates in the present tense, but resurrection attacks exploit permanent anchors – the memory of a loved one. A fake video of a dead grandmother endorsing a political candidate cannot be debunked with a tweet; her loved ones are ethically paralysed. Should they denounce the video, they may insult her legacy. Should they stay silent, the propaganda stands. British counterintelligence has already logged 43 such attacks since last June, targeting voters in marginal constituencies.
The government’s response has been characteristically cautious. Civil liberties groups remind us that any censorship framework for deepfakes could also suppress legitimate satire or historical re-enactments. And yet, the NCSC has quietly begun labelling certain AI-generated content as ‘necrobotic’ in internal memos, a term that hints at a new classification of hybrid threat. Private sector platforms are being pressed to adopt cryptographic watermarking of training data, so that any resurrected avatar can be traced to a specific original source. But as one GCHQ analyst put it: “Tracing a ghost is like mapping fog.”
For the average citizen, the first line of defence is digital hygiene. The recommendation is now to limit public sharing of high-resolution photos and voice samples, especially of elderly or deceased relatives. Some bereavement charities have started advising families to post obituaries without photos. This feels draconian, but the alternative is crueller – having your loss repurposed as a vector of division.
We are witnessing a perversion of one of our oldest human hopes: to speak with the dead. Currently it is deployed for tactical advantage, but the same tools could be used to comfort, to preserve cultural heritage, or even to solve cold cases. The line between memorialising and manipulating is being drawn with machine precision. Britain’s spies have warned us – now it is up to society to decide whether we let code dictate how we mourn, and how we trust.
