The Five Most Key Takeaways from This Blog
- A summer 2024 revelation that a hacker got access to OpenAI’s ChatGPT internal messaging in early 2023, raising relevant questions about how secure OpenAI’s A.I. technologies really are.
- Given the secrecy surrounding this hack (OpenAI did not choose to publicly disclose it at the time, given that no customer data was stolen), it could lead one to wonder what other internal security problems, or vulnerabilities, may be unknown.
- This revelation comes as OpenAI has secured a major deal with Apple to have OpenAI’s A.I. technologies powering certain features in upcoming generations of Apple products.
- Many individuals worry about the vulnerability of their data. But in the context of business, a data breach that exposes data like employees’ Social Security numbers and customers’ credit card numbers can be disastrous.
- For business owners wondering what other options are available for A.I. solutions like customer-service chatbots, the long-popular IBM Watson stands as a viable option.
Keeping Data Safe
As a business owner, you may have experimented with using ChatGPT a handful of times.
Or you may already be using it on a regular basis. So many emails, so little time, so thank a deity or deities for chatbots.
However, you ought to think twice before using ChatGPT for content that relates to your customers, or internal operations for that matter.
This is because data-privacy concerns about ChatGPT are beginning to grow.
Notable Data-privacy Moments in ChatGPT History: Bans and Hacks
The early-2023 hack has raised concerns both inside and outside the company about the vulnerability of ChatGPT.
In a separate incident, OpenAI did admit in March 2023 that the company had to shut down service because a bug made it possible for user data to be accessed.
So even before the 2024 public reveal of the hack, there were some notable high-profile expressions of data-privacy worries.
One such worry came from the nation of Italy in the summer of 2023. Italy went so far as to ban ChatGPT within the country because of concerns that ChatGPT actively stores users’ information.
That, and because Italy also had concerns about the lack of age verification, which makes it easy for children to access inappropriate ChatGPT-written content.
The Italian government has since lifted the ban, but controversy and concerns about OpenAI’s data-privacy vulnerabilities still remain.
But at the same time, multibillion-dollar backing from companies like Microsoft and Apple is more than enough fuel for keeping the A.I. show on the road.
So within the United States at least, you can expect to see ChatGPT (and DALL-E) keep on truckin’.
Should You Be Concerned About ChatGPT and Data Privacy?
The short answer is: probably, yes.
A major revelation is that ChatGPT indeed stores and collects personal data given to it.
This ought to be concerning for anyone who uses ChatGPT, because its range of questions that it is allowed to answer is very wide. Therefore, its range of answers is similarly wide.
Do note, however, that OpenAI says that it does not train its systems on data from ChatGPT Enterprise, Chat GPT Team, or API users. But training is a different thing from storing, which OpenAI apparently still does even for Enterprise users, as seen on this FAQ page:
Within your organization, end users can view their own conversations. Your organization has control over workspaces, and workspace admins can access an audit log of conversations and GPTs through the Enterprise Compliance API . Authorized OpenAI employees will only ever access your conversations for the purposes of resolving incidents, recovering end user conversations with your explicit permission, or where required by applicable law.
So if another major data breach happens, though, you might see a spreading of personal data throughout the Internet.
However, what we have seen happen is that ChatGPT can suffer pretty serious data breaches, which is something to be worried about. This is especially true given its ever-growing popularity, making it a big target for cybercriminals.
IBM Watson for Customer Outreach
For a safer alternative to ChatGPT, go with IBM Watson.
It is one of the most time-tested chatbots out there, though it is important to understand that it does not do everything that ChatGPT does.
Specifically, what IBM Watson does not currently do is ask-me-anything writing assignments.
Though Watson will not spit out a memo for you in record time, it does have the ability to have safer, more secure conversations directly with your customers.
And since IBM Watson has been tested by countless companies across a wide variety of industries for more than a decade now, you can rest assured that IBM Watson has a much lower chance of “hallucinating” while talking to your customers.
So, that is a big relief, being able to employ a chatbot that you do not need to monitor in its direct customer interactions.
IBM Watson is trained on your company data, so that it knows what customers will be asking and what information is important to know about your products and services.
Recent Comments