Office management edition
A familiar face is no longer proof. Meta’s new model makes mistakes with payments, and Azure is getting more expensive in Europe.
Interpol: spot the pressure and check requests for money
Police: scammers use AI to impersonate colleagues and family. Always check requests for money.
In a campaign by Interpol, police warn about scammers using AI to impersonate people you know. They copy the faces and voices of family members, friends and colleagues. This kind of fake is called a deepfake. It is an image, video or audio recording that has been made or changed using AI. It can make it seem as if a real person is speaking. Scammers need very little material from online sources to do this.
Police give three examples. A ‘colleague’ calls by video and asks for money to be transferred quickly. A family member calls in an emergency. Or someone you know online asks for money. The campaign’s message is: pause, check and do not fall for it. A familiar face or voice does not prove who someone is.
A deepfake relies on two things: a convincing face or voice, and pressure to act quickly. So pay attention to behaviour. Be alert if someone is rushing you, claims there is an emergency or tries to frighten you. A request to keep something secret is also suspicious. The same applies to an unusual way of paying. Does someone say you must not discuss it with anyone else? That is suspicious too. It takes away the very check that could help you.
The advice is: hang up and call the person back on a number you trust. Do not use the details given by the caller. Police advise businesses to follow set steps when checking payments. This applies even if the request seems to come from a trusted source. Police say this type of scam is hard to spot. Faces and voices can seem real. So simply looking and listening is not enough to check.
Background for subscribersPolice warn that AI can convincingly copy faces and voices. So check an unexpected request for money by calling a number you already trust. In businesses, it helps to agree on set steps for checking payments. If you see AI features in an accounting tool or on a work phone, ask what they are used for and what rules apply. New features are sometimes made available in stages. So also check whether your organisation has access to them.
Meta’s Muse Wonder is strong at reports, but makes mistakes with payments
Meta released Muse Wonder on 12 October. The company calls it its strongest model so far. It scored 83.2 per cent on DualEntry’s accounting benchmark. A benchmark is a set of tasks used to compare models. With this score, Muse Wonder shares second place among 57 AI models tested. Its predecessor, Muse Spark 1.1, scored 61.4 per cent. Only Grok 4.5 scored higher, with 84.2 per cent. You can use Muse Wonder through the Meta Model API and Muse Code. Muse Wonder Full is available only with two Muse Code subscriptions: Power (50 dollars per month) and Max (200 dollars per month). The Muse app will get the model ‘soon’. No date has been given. When asked questions about reporting, Muse Wonder gave the right answer almost every time in repeated tests. These included cash flow and calculations for annual accounts. Its results were less consistent in everyday work with transactions. The model sometimes made mistakes when paying outstanding invoices. It also did not always make the same entries at the end of the month. This happened with difficult cases too, where transactions had to be classified. So a model can be good at making a summary and still make mistakes in individual entries. The errors were in this everyday transaction work. There was also no safety report at launch.
Background for subscribersThis is a taster.
Subscribers read the whole Office management edition: every story, each with its background in plain language.
Subscribe →Not ready for that? Read the free letter first →
AI and tech, made simple.
This is the Office management edition. Subscribers get the background to every story, in plain language.
Subscribe
inklaretaal · Amsterdam · © 2026