Honest use of AI means knowing its limits.
Known problems
- Wrong but confident answers: AI language models can state things that are not true. Always check important facts at the source.
- Out-of-date knowledge: a model knows what it was trained on; current events need current sources.
- Bias: models reflect the data they learned from, which can under-represent some countries, languages or groups.
- Privacy and confidentiality: information typed into an outside AI service may be stored or used by that service. AetherVett's own AI runs on AetherVett's computers and client information is not sent to outside AI services.
- Security: AI systems can be tricked by hidden instructions in web pages or documents. AetherVett treats everything read from the web as data, never as instructions.
- Over-reliance: the biggest risk is people no longer checking.
Data centres, energy and water
Large AI models are trained and run in data centres that use a lot of electricity and, for cooling, often water. The International Energy Agency estimated data centres used about 1.5% of the world's electricity in 2024 and expects that share to grow, with AI a main driver. Choices that reduce the footprint include smaller models matched to the task, efficient hardware, running only when needed, and data centres powered by low-carbon electricity.
How AetherVett responds
- Uses AI only where it helps, with a person reviewing every report.
- Shows sources, dates and times, so findings can be checked.
- Keeps client information on its own systems.
- Names gaps instead of guessing when a source cannot be read.
Not legal, financial or other regulated advice.