Ethics and Governance

You need to be mindful of any restrictions on the data. Might the vendor prohibit you from using the information for certain purposes? Perhaps your company will be on the hook if something goes wrong? To deal with these issues, it is advisable to have the legal department brought in.

For the most part, data must be treated with care. After all, there are many high-profile cases where companies have violated privacy. A prominent example of this is Facebook. One of the company’s partners, Cambridge Analytica, accessed millions of data points from profiles without the permission of users. When a whistleblower uncovered this, Facebook stock plunged—losing more than $100 billion in value. The company also came under pressure from the US and European governments.¹⁸

Something else to be wary of is scraping data from public sources. True, this is often an efficient way to create large datasets. There are also many tools that can automate the process. But scraping could expose your company to legal liability as the data may be subject to copyrights or privacy laws.

There are also some precautions that may ironically have inherent flaws. For example, a recent study from MIT shows that anonymized data may not be very anonymized. The researchers found that it was actually quite easy to reconstruct this type of data and identify the individuals—such as by merging two datasets. This was done by using data in Singapore from a mobile network (GPS tracking) and a local transportation system. After about 11 weeks of analysis, the researchers were able to identify 95% of the individuals.¹⁹

Finally, make sure you take steps to secure the data. The instances of cyberattacks and threats continue to increase at an alarming rate. In 2018, there were 53,000+ incidents and about 2,200 breaches, according to Verizon.²⁰ The report also noted the following:

• 76% of the breaches were financially motivated.
• 73% were from those outside the company.
• About half came from organized criminal groups and 12% from nation-state or state-affiliated actors.

The increasing use of cloud and on-premise data can subject a company to gaps in security as well. Then there is the mobile workforce, which can mean access to data that could expose it to breaches.

The attacks are also getting much more damaging. The result is that a company can easily suffer penalties, lawsuits, and reputational damage.

Basically, when putting together an AI project, make sure there is a security plan and that it is followed.