Your AI hiring tool might be high-risk under the AI Act. Here's what that means.
AI Act transparency rules apply from August 2, 2026. High-risk employment AI rules apply from December 2, 2027. Practical preparation guide.
Correct language and a familiar voice do not establish identity. Verify bank-detail changes through a known contact and rehearse the team response.
A supplier's email asks you to change its bank details. The tone is familiar. Verify the request through a known contact before paying. FBI IC3 recommends a secondary channel for account changes, because the message alone cannot establish authority.
An attacker can draw on company websites, social posts or earlier correspondence. Generative tools can help with tone and language. The text may not reveal whether AI was involved at all.
Do not make spelling your identity check. Inspect the address, unusual requests and changed payment details. A legitimate account can also be compromised, so a sensitive operation needs a step independent of the received message.
A caller may demand an urgent transfer, replacement bank details or credentials. You do not need to prove the voice is synthetic before pausing the operation. Call back using the known number in a contract or directory rather than a contact supplied in the suspicious request.
Give bank-detail changes an independent approver. The procedure should work whether or not the attacker used AI.
Technology alone isn't enough. Write the procedure down and rehearse it with the people who will use it.
A request that changes banking details should trigger verification through a second channel, and the change belongs in an audit log. The rule should be evaluated from confirmed incidents and false positives, not an unsupported marketing anecdote.
Prioritise missing controls by risk and name an owner for the exercise. Test the reporting route and the response. Our security guidance covers access questions. We can help design the exercise.
AI Act transparency rules apply from August 2, 2026. High-risk employment AI rules apply from December 2, 2027. Practical preparation guide.

Choose a first automation by scoring the work, data, risk, ownership, and reversibility. Then validate one pilot before you expand it.
A model plan for one recurring task. Measure today’s work, set data permissions and output review, then decide whether the trial should continue.