A new guest has stormed into the offices of consulting firms. It’s called artificial intelligence. And his ability to sneak into any report, advice, or work done by his employees is already raising concerns at the top of professional services firms. This is demonstrated by an internal email sent by Deloitte Spain on April 11, to which this newspaper has had access. “Confidential information owned by the client, third parties or Deloitte should not be contributed to these AI platforms,” says a sentence of the four-page text sent to the company’s workers. In it, the company, with more than 8,000 professionals in Spain, warns that any use of ChatGPT or similar tools must first have their approval.
The speed with which the use of this disturbing technology is taking hold in large corporations. The Australian division of another Big Four (as the four major global consultancies are known), PwC, already sent a message last February asking its employees not to use material created by these applications with clients. In the case of Deloitte Spain, the fears are many. Unauthorized information coming to light: “If the data entered is stored on a server or database, there is a potential risk that it will be disclosed to third parties without authorization, either due to security failures, cyberattacks, or violations of data,” the email says. To the fact that there are biases: “If the training data used contains biases or prejudices, there is a risk that the responses generated by the model will be discriminatory or unfair, even affecting the privacy of individuals who interact with the model and the reputation of the companies that use it. Or that they are taken for good answers riddled with errors: “There are limitations in the ability of these tools to understand the context and meaning behind the words. This forces the implementation of adequate validation and human supervision mechanisms”.
Mistrust about the usefulness of the information flies over the entire document. “These tools are trained using historical data, which means they may not be suitable for spotting new trends or issues. […] It may entail a risk of lack of rigor and precision in the work”.
Consultants are highly requested by governments, multinationals and other entities in need of advice with all kinds of programs and spending plans. In such delicate schedules, highly dependent on technical knowledge that in many cases is now available in ChatGPT, there are workers who may be tempted to resort to artificial intelligence as the fastest way to resolve their doubts. Deloitte does not believe this is ideal, both for the reasons listed above and others. “If the responses generated contain sensitive, protected (for example, copyrighted) or confidential information, there is a risk that this information will be used inappropriately,” she points out.
Client’s aprovement
The casuistry in which they can provoke the anger of those who hire them is extensive. “If Deloitte uses these tools to obtain information about clients or to make decisions about them without their knowledge or consent, this may be considered unethical and could damage our reputation,” they warn. They also consider dangerous the use of artificial intelligence “to obtain information about customers or to make decisions about them without their knowledge or consent.”
The firm asks its employees not to use their corporate emails to open accounts in OpenAI —creator of ChatGPT— unless it has been expressly authorized. And if they have already opened it, he urges them to notify their manager “as soon as possible.” In addition, they must inform the client if they use AI in a project “and obtain their acceptance”.
Suspicions about bad practices that may arise from the use of artificial intelligence are not exclusive to companies. The European Union wants to regulate ChatGPT because of its effects on user privacy, its potential to spread misinformation, and possible job destruction. And one of its main members, Italy, has gone even further and has chosen to block access for breaching data protection regulations, considering that it collects user data illegally.
Follow all the information of Economy and Business in Facebook and Twitteror in our weekly newsletter
Five Days agenda
The most important economic appointments of the day, with the keys and the context to understand their scope.
RECEIVE IT IN YOUR MAIL
#Deloitte #asks #employees #Spain #share #confidential #information #ChatGPT