Corsight opens the black box for AI explainability

Facial recognition company Corsight AI is working on new research and platforms that can advance AI explainability and transparency.

A part of the initiative is visually demonstrating which parts of the face an AI software is analyzing to find similarities. During a lineup of police suspects, the system can find look-alikes of a suspect and demonstrate how and why it differentiates between them, according to the company’s Vice President of Research and Development Matan Noga.

“This is the essence of explainability for us: By tapping into what previously was considered the black box and presenting it in a human way we can better understand and justify what the system is doing,” says Noga.

Noga dived into the topic at a webinar on the legal and commercial implications of AI explainability hosted by Shlomit Yanisky-Ravid, a professor of Intellectual Property (IP) Law at Yale and Fordham University.

AI explainability ensures that users can understand why a particular outcome was reached, ensuring both trust and regulatory compliance. To ensure AI management systems are explainable and ethical, the International Standards Organization issued the ISO 42001 standard in 2023 with Corsight AI becoming one of the first companies to obtain the certificate.

The Tel Aviv-based firm, however, has also been implicated in Israel’s surveillance of Palestinians.

While questions about their work in Israel did not come up during the webinar, Corsight AI said that it is solving questions related to accountability by providing its clients, including those from the government and retail sector, legal support and even ethical recommendations related to its products.

“We are accountable for what the system is recommending but it is the actual customer, whether it’s an agency or retailer, who is responsible for their actions using the system’s recommendations,” adds Noga.

Priorities related to AI explainability among clients can be different, depending on their geographical location and political winds. In Europe and the UK both government and commercial clients generally tend to prioritize privacy and fair use under the influence of the AI Act. In Latin America where violent crime is widespread, public opinion tends to prioritize personal safety over privacy, Noga said during the webinar.

Meanwhile, Corsight AI is trying to appeal to other markets that are keen to crack down on crime.

During the Future Crime Summit 2025 in New Delhi, India in February, Amita Singh, a board member at security company I-Sec, spoke about the company’s capabilities in surveillance.

“While digital forensics helps analyze data, Corsight AI’s facial intelligence technology provides tangible, real-world evidence by identifying individuals in surveillance footage – even in poor lighting or when they attempt to obscure their features,” says Amita Singh.

In 2021, Corsight AI signed a Memorandum of Understanding (MoU) with I-Sec and Amtron, an agency implementing IT projects in the Indian State of Assam, to establish a facial recognition excellence center in Guhawati.

Article Topics

AI  |  biometric identification  |  biometrics  |  Corsight  |  criminal ID  |  ethics  |  explainability  |  facial recognition  |  forensics