May 16, 2019

Is AI compatible with a Right to Privacy?

Navigating the right to privacy in the digital age, and especially the artificial intelligence age: AI cannot function without data, and AI that is helpful to people on a personal basis needs to be based on (i.e. "trained on") personal data.

Is AI compatible with a Right to Privacy?

In many nations, including the US and China, the right to privacy is enshrined in the constitution governing the country, and refers to the right to have a safe personal space in the home which is protected from incursion or intrusion. These terms provide a basis of what is referred to as a right to privacy, and in both cases above there are provisions to break through the privacy barrier in the case of law enforcement:

The right of the people to be secure in their persons, houses, papers, and effects, against unreasonable searches and seizures, shall not be violated, and no Warrants shall issue, but upon probable cause, supported by Oath or affirmation, and particularly describing the place to be searched, and the persons or things to be seized. - Amendment 4, US Constitution, 1789
The home of citizens of the People's Republic of China is inviolable. Unlawful search of, or intrusion into, a citizen's home is prohibited. The freedom and privacy of correspondence of citizens of the People's Republic of China are protected by law. No organization or individual may, on any ground, infringe upon the freedom and privacy of citizens' correspondence except in cases where, to meet the needs of state security or of investigation into criminal offenses, public security or procuratorial organs are permitted to censor correspondence in accordance with procedures prescribed by law. - Article 2, Chapter 39-40, China Constitution, 1982/2004

Understandably, the governing constitutions legally refer to the rights between a person and a government, rather than between a person and corporations the person may be doing business with, but the important thing is that the idea of a right to privacy is a fundamental idea in society. Navigating the right to privacy in the digital age, and especially the artificial intelligence age, is another question. AI cannot function without data, and AI that is helpful to people on a personal basis needs to be based on (i.e. "trained on") personal data.

There are three broad scenarios in terms of how personal data is gathered and used in AI systems. It is useful to consider the implications of each of them.

AI-enabled Products

These are products that a consumer uses that directly implement AI-enabled features, such as the iPhone and Facebook. It is clear that a user contributing usage data and content into the company's database, and it can be reasonably assumed that the company can leverage the data over time to improve the product. Each company may differ on how personal data is used and aggregated. Apple famously tries to keep the user's data private and available only to the user himself or herself and not available to Apple for further use, while Facebook will use data that users have entered to train new algorithms for the benefit of its whole population. It is a reasonable assumption that most tech companies are in the process of integrating artificial intelligence into their products today.

Data Brokers

You may be surprised at how many traditional business models are based around buying your data from companies and selling them to others, but these include data such as credit reporting, advertising and purchasing data and personal information. Painting a more complete picture of a person allows a company to better sell new products to that person.

By buying or licensing data or scraping public records, third-party data companies can assemble thousands of attributes each for billions of people. For decades, companies could buy up lists of magazines subscribers to build targeted advertising audiences. These days, if you use a smartphone or a credit card, it’s not difficult for a company to determine if you’ve just gone through a break-up, if you’re pregnant or trying to lose weight, whether you’re an extrovert, what medicine you take, where you’ve been, and even how you swipe and tap on your smartphone. - Fast Company

Public Data

These are products built from data from the "commons" or the public, including security cameras, police footage, public spaces, traffic data. Generally they are collected using resources spent by the government and funded by taxpayers. Public data may be freely available to download, usually in an anonymized fashion, or they may be provided to third parties who are being contracted to do analysis or develop new services. This includes new sensors that cities are deploying for "smart city" initiatives. But given the government is funded by citizens, they arguably have a say in how that data may be used.

The pace of technology development greatly outpaces that of legal responsibility, so there is a gray area where it is unclear where the right to privacy meets the acceptable use of data gathered by myriad data collection systems. It is important for each person understand the relationships between all of the parties as well as the potential uses in AI and data-driven services so that, as a society, we can determine the appropriate balance of what is acceptable. After all, we have a longstanding right to privacy – we just have to figure out how it applies in today's world.