There is a greater than 10% chance we are all overreacting

News outlets last week were filled with figures citing a ten percent chance that artificial intelligence will end humanity. Emergency committee sessions are meeting in parliament, while executives at OpenAI and Anthropic, alongside researchers like Geoffrey Hinton, debate the risks of autonomous systems.

It makes for dramatic reading. It also misrepresents how engineering works in practice.

Software does not develop personal motives. It does not wake up, acquire ambition, or decide to undermine its user. When technology causes harm, the failure sits squarely with human intent, poor configuration, or a lack of oversight. The risk stems from a human misusing a powerful tool, not the tool acting on its own desires.

Technology operates in a continuous balance between risk and protection. Every advance that creates a vulnerability triggers a counter-advance designed to contain it. Cyber security exists because open networks exist. Anti-virus tools exist because malware exists. The balance between capability and defence is standard engineering, and artificial intelligence follows that same path.

Sensational headlines about rogue AI serve a convenient purpose for technology suppliers. Framing the issue as an uncontrollable entity shifts attention away from basic operational discipline. When the risk is treated as an abstract monster, nobody asks hard questions about testing, monitoring, or deployment standards.

Underneath the headlines, AI runs on software engines following strict parameters. The reason AI deployments break inside accountancy, legal, or recruitment firms is rarely a technical rebellion. It breaks when systems are deployed as unmanaged tools, handed over without defined parameters, edge-case handling, or daily maintenance.

You do not need to worry about an algorithm taking over your firm. The operational reality is much simpler: software follows its configuration, and reliability requires ongoing management, clear role specifications, and strict oversight. The headlines will take care of themselves.