It can involve impersonation or misleading communications across channels such as messages, calls and video. Generative systems may change the scale or apparent realism of those communications without changing the underlying exploitation of trust.
Defensive assessment should distinguish warning signs from confirmed compromise. Verification, appropriate access controls and coordinated response matter, while one suspicious message does not establish who acted or whether an account was successfully compromised.
ELI5
Social engineering is trying to persuade someone to trust a misleading request rather than directly breaking a technical lock.
For example, a message may appear to come from a colleague but ask for an unusual disclosure. The important defensive step is to verify the request through an appropriate trusted process, not assume appearance proves identity.
Does a realistic voice prove identity?
No. Apparent realism alone does not establish who is communicating or whether the request is authorized.
Is one warning signal proof of compromise?
No. Signals support investigation and should not be converted into unsupported claims about a person or account.
