It seems like these days everyone is getting in on the artificial intelligence act. But there are also many people who refuse to go near it.
This could be because they are concerned by what AI chatbots can do. There have been many movies, shows, and scare stories in the news about AI becoming sentient and leading us humans all to Armageddon.
So, should you be cautious when using AI services, such as chatbots and text-to-image generators? Is ChatGPT safe to use? In this blog post, we aim to fully answer this question so that you can put your mind to rest.
Is ChatGPT Safe?
When it comes to interacting with technology, safety is a complicated issue. As for AI, it can really go all over the place. There are prominent AI-doomers, including Nick Bostrom and Eliezer Yudkowsky, who fear that AI will soon wipe all human life off the earth. They’re not kidding, and it is a non-zero concern that very few people are talking about.
But there are many more who worry about… less existential risks, like will ChatGPT steal my credit card number or password? or will ChatGPT give me a completely wrong answer?
To the average person, these are much more significant concerns, as there’s not really much an individual can do to stop the machines from taking over if they really want to.
Content Safety
Even when you ask the developers who are in charge of training AI, many will say that when using these newfangled tools, you need to be aware of what you’re asking for. And that’s because chatbots like ChatGPT definitely have the ability to generate content that can be both biased and harmful to others.
Famously, an earlier LLM known as Tay, developed by Microsoft, began spitting out racist commentary very quickly after its introduction.
Even when you ask the developers who are in charge of training AI, many will say that when using these newfangled tools, you need to be aware of what you’re asking for. And that’s because chatbots like ChatGPT do have the ability to generate content that can be both biased and harmful to others.
Content Errors and Hallucination
Whenever you use ChatGPT for serious work, such as school work, legal work, website scraping, or work with databases and tables, you should always proceed with caution. This is because it is known to sometimes generate text that contains multiple errors, and can be factually and totally incorrect.
This can lead some people to spread incorrect information that they shouldn’t have trusted.
A famous example is the lawyer who submitted a brief to a judge, written by ChatGPT, that contained references to legal cases that didn’t exist. Oops. And maybe loss of a law license!
Proceed With Caution
Because AI is such an emerging technology, it is difficult to give a straight answer when it comes to whether or not ChatGPT is safe to use or not. However, like with anything that is new and unregulated, it’s probably best to proceed with caution.
ChatGPT’s parent company, OpenAI, isn’t necessarily the bad guy here either! Because although its new technology is constantly updated, bringing even more capability to the AI model, the increasing demand means that there is an increased risk of scamming, hacking, and cyber attacks.
If you choose to pay for a ChatGPT subscription, then you will have access to faster speeds, and most importantly, a more accurate chat bot! With all the new features, you may get more peace of mind when it comes to safer use.
Paving The Way For Transhumanism or Cyber Crime?
Many users find that ChatGPT is a very powerful tool that also demonstrates how quickly artificial intelligence is paving the way for a future where we’ll be having human-like conversations with chatbots and voice bots.
But this also makes some people very wary and fearful as to what AI can potentially do, especially when it is being misused by humans to do things like hacking, getting ChatGPT to create malicious code for cybersecurity exploitation and other cyber criminal activities.
In addition, it has recently been revealed that OpenAI has been using contractors to build ChatGPT’s artificial intelligence. And it’s this gig work economy that could potentially crack holes in this AI matrix, because ‘Advanced AI Data Trainers’ make on average about $15 an hour.
Therefore, it’s not at all difficult to imagine a disgruntled worker going postal, and making the AI chatbot go rogue to do their personal bidding.
Or even imagine them helping to get ChatGPT to respond with misinformation and propaganda when triggered with certain words, elaborating on the way that it hallucinates now, when it generates text that doesn’t make sense or is completely fabricated and wrong.
And although OpenAI has, arguably, some of the most robust cybersecurity measures in place, there’s always the risk that details of your account will be exploited. That’s because OpenAI not only has access to all of your account details (and payment data if you’re a ChatGPT-Plus subscriber) but to all of the information that you choose to share with ChatGPT.
ALL OF IT.
So, if you choose to trust a company that pays $15 an hour to workers that are helping to build the AI, that’s your choice. In fact, in one incident in March 2023, which saw ChatGPT’s service go down for several hours, some users of the AI chatbot could view other people’s chat history, which could have included personal information that they didn’t want to be seen.
Plus, even more concerning, was the fact that subscribers of ChatGPT-Plus may have had their payment information leaked during the incident, which is highly concerning for a new service where everyone is jumping on the bandwagon, but which there is no real track record of transparency and good governance.
Cybersecurity experts have also expressed concern about ChatGPT’s potential as a tool for hackers to use for creating and running dark web forums, along with well written email phishing scams (and future AI-voice cloning phone calls and deepfake video) that will fool many unsuspecting people.
One recent example of this happened when someone posted on their Twitter that they had asked the AI chatbot to provide instructions on how to hack computers. The details generated by ChatGPT on how to do this sent shivers up many spines.
Can Hackers Make ChatGPT Unsafe?
ChatGPT is available to use by anyone, which means that not everyone that uses it will have the best of intentions. Therefore, hackers, terrorists, enemy governments, etc. are able to access ChatGPT to help them with their potentially dangerous endeavors.
For example, ChatGPT, when correctly prompted, is able to generate computer code that can be used to create hacking programs that can cause a lot of damage if they are let loose from the wrong hands and lead to cyberattacks and other criminal activity.
ChatGPT, when correctly prompted, is able to generate computer code that can be used to create hacking programs that can cause a lot of damage if they are let loose from the wrong hands and lead to cyberattacks and other criminal activity.
Even recently, there have been quite a few Twitter threads like this one that warn even the latest iteration of the AI chatbot, Chat GPT-4, is still vulnerable to jailbreaks like the (fairly amusing) DAN, and what parent company OpenAI calls ‘prompt injection attacks’.
In addition, ChatGPT’s new plug-ins is even able to run self-generated code. And although OpenAI has fenced in this capability, the GPT-3 application programming interface has already been hacked, and this scenario can definitely happen again once more and more people start using the AI chatbot.
Deleting Your Account If Necessary
Luckily, there is something easy that you can do after using ChatGPT that will allow you to delete your account, if you’re concerned about how it shares your information and data. Basically, what you need to do is sign into ChatGPT and make your way to the Settings tab. In the popup, you will see Data Controls.
Click on Show. At the bottom of the Data Controls menu, you’ll find Export Data and Delete Account.
In order to see what data you’ve shared with the AI chatbot, you can click on Export Data. If you want to close your account, then click on Delete Account. Another popup will appear allowing you to click on the red Permanently Delete My Account button.
Remove Individual Conversations
If you don’t want to take the more drastic option of permanently deleting your entire account, then ChatGPT also gives you the ability to get rid of specific conversations that you have with the AI chatbot. Simply click on the Clear Conversations tab and click through the options.
If you don’t want to take the more drastic option of permanently deleting your entire account, then ChatGPT also gives you the ability to get rid of specific conversations that you have with the AI chatbot. Simply click on the Clear Conversations tab and click through the options.
Request to Opt Out
Alternatively, if you want to continue using ChatGPT, but don’t want to close your account or don’t have the time to manually delete certain conversations, then OpenAI also gives you a way to do this. There is a User Content Opt Out Request that you can fill out.
However, this option is technically only a request. And OpenAI will only consider your request if you pay for a ChatGPT-Plus subscription. You should also take note that you won’t be able to unlink your phone number, even if you delete your account permanently, due to security and anti-fraud reasons.
Final Thoughts
So, is ChatGPT safe? Well, there is no clear yes or no answer for that question. That means it will all depend on how much of a risk you want to take by knowing that artificial intelligence probably already knows more about you than you will ever know about it.
