The survival guide for people who want to understand the frightening scenarios without being frightened into nonsense.
Dramatic question in. Evidence, uncertainty and practical action out.
Start with the risk that worries you most. Each section separates what is plausible, what is uncertain and what you can do about it.
What would an AI takeover actually require? Separate plausible pathways, known constraints and science-fiction assumptions.
Explore cluster →WORK & INCOMETask-level exposure, career resilience and what AI can automate without pretending anyone can predict your redundancy date.
Explore cluster →FRAUD & VERIFICATIONVoice cloning, deepfake video, phishing and the verification habits that reduce your odds of being fooled.
Explore cluster →DIGITAL RESILIENCEWhat AI changes about cyber risk, plus practical ways households and small organisations can remove single points of failure.
Explore cluster →PREP WITHOUT PANICUseful preparations for outages, disruption and misinformation that remain sensible even if the AI apocalypse never arrives.
Explore cluster →IDENTITY & PRIVACYPasskeys, identity theft, voice cloning and practical ways to reduce the personal data attackers can turn into convincing AI impersonation.
Explore cluster →VERIFY BEFORE YOU BELIEVEHow to verify AI-generated images, video, text and viral claims using provenance, source checks and repeatable workflows rather than detector hype.
Explore cluster →WHO SETS THE RULES?Who regulates AI, what governments and companies can actually control, and where accountability sits when automated systems cause harm.
Explore cluster →WORKPLACE AIConfidential data, employee use, AI policies and practical rules for using generative AI at work without creating avoidable privacy or security problems.
Explore cluster →EVIDENCE DESKScheming, alignment, agent incidents and the difference between a controlled safety evaluation and a demonstrated real-world capability.
Explore cluster →
TAKEOVER & EXTINCTIONAI models can output false statements accidentally, and some frontier models have also shown strategically deceptive behaviour in controlled evaluations. That is a safety concern, but it is not evidence of human-like motives.
Read guide →
WORK & INCOMEAI is more likely to change bundles of tasks than erase every person with the same job title at once. Your exposure depends on how much of your week is digital, repetitive and checkable versus physical, relationship-heavy, ambiguous or accountable.
Read guide →
FRAUD & VERIFICATIONYou do not need to win a forensic argument about whether media is fake before protecting yourself. Pause the transaction, verify the person through a separate route and confirm any payment or account change using established details.
Read guide →
PREP WITHOUT PANICThe sensible AI survival plan looks remarkably ordinary: resilient income, protected identity, reliable backups, a household communication plan and enough basic supplies to ride out short disruptions.
Read guide →
EVIDENCE DESK“Scheming” research tests whether a model can pursue an objective covertly while misleading oversight. Frontier models have shown such behaviour in deliberately constructed evaluations, which is a reason to improve monitoring — not proof that ordinary chatbots are secretly plotting.
Read guide →
DIGITAL RESILIENCEAn autonomous cyber agent can execute a sequence of attack tasks with less step-by-step human input. That raises speed and scale concerns, but the agent still needs reachable targets, tools, credentials or exploitable weaknesses.
Read guide →Score the tasks you actually do, see which parts of your role are most exposed to AI, and avoid relying on simplistic job-title predictions.
Use it →FREE TOOLCheck the process around a suspicious call, video or message. This does not claim to detect synthetic media.
Use it →FREE TOOLFind the single points of failure in your household or small business before an outage finds them for you.
Use it →AI systems are becoming more capable and more connected to real tools. Researchers are actively testing deception, agent autonomy and cyber risk. Those are serious subjects. They become less useful when every uncertainty is converted into a certainty for clicks.
So the editorial deal is simple: we can use “AI doomsday” in the headline; we cannot manufacture a doomsday underneath it.
Read our editorial standard →