Original text here from Patrice Bernard (LinkedIn)
Combating fraud is one of the most popular applications of artificial intelligence. With many scams starting with a phone call, Microsoft offers telecommunications operators a spy software that detects them in conversations. As the English saying goes: "what could go wrong?".
The case for implementation is clear and particularly appealing to industries most at risk, including the financial sector: hearing the stories of people deceived by a scammer on the phone, those with a slight understanding of the subject often despair at missing clear warning signs before the trap snaps shut. So much so that employing AI to identify these clues and sound the alarm under such circumstances seems almost a given.
Microsoft's implementation of this principle starts "simply" by monitoring telephone exchanges through the operator's network, to which the solution is connected and hosted on Microsoft's Azure cloud. When the real-time data analysis models suspect an ongoing scam, they send a warning SMS to the potential victim, followed by a summary at the end of the call explaining the risk and precautions to take.
Subscribe to our newsletter:
Currently, this system is deployed as a free pre-version, with no pricing details available yet. However, the technical integration required is likely not as straightforward as official communications might suggest. This phase is probably about validating large-scale functionality and gauging public reception.
Naturally, safeguards are in place to ensure privacy. Besides Microsoft's security infrastructure promises and the assurance that the content of conversations is never stored, activation is supposed to occur only at the explicit request of the end-user (without which, the service would presumably not be legal, even in the United States).
While the concept is inherently attractive and leaves little doubt about its potential effectiveness in fraud prevention, the method used undoubtedly opens a Pandora's box. Regardless of arguments provided to alleviate concerns, opening access to telephone networks inevitably creates a vulnerability that could be exploited for malicious purposes, including mass surveillance. In my opinion, no goal justifies such a risk.
