Why Thanking AI Doesn’t Matter – Smarter Ways to Stay Secure Online

When ChatGPT launched, most of us instinctively added a polite "please" or "thank you" to our conversations. Like Sheldon Cooper, some of us are polite to ensure the AI knows that we've always been on their side when the robot uprising finally arrives.

But here's the thing: AI doesn’t think, feel, or care about your mood.

When we take those extra few seconds to type "please" and "thank you," we're projecting human traits onto a stylishly shaped collection of silicon and other materials, operated by a series of zeroes and ones.

And, according to AI companies, we're costing them millions of unnecessary dollars.

Anthropomorphizing is a Human Habit

Even if it feels awkward not to be polite, the truth is that AI does not "reason" or "understand". While using a courteous tone with AI may prompt polite answers, others believe it's merely a form of self-soothing. It helps us humans make sense of the complex tech behind our newest tech tool.

"Anthropomorphism" means projecting human characteristics to non-human entities. When we do it with AI tools, we may be creating false expectations of AI's capabilities. We're stubbornly insisting that it has some qualities a human might have.

That can be dangerous.

Last year, a teenager became a victim of a companionship algorithm. He had such an intense, emotionally charged relationship with a Replica.ai chatbot named Danaerys that he committed suicide to 'join her.'

The Good, the Bad, and the Ugly of AI

AI is not inherently good or bad. It gets shaped (trained) by humans who display the gamut of human traits from "practically angelic" to "rotten to the core".

Some AI tools are designed with helpful goals in mind to tackle a wide range of queries and provide users with detailed responses to their questions. But others are crafted for harm, such as those used to generate malware or manipulate people with social engineering tactics.

There are AI models capable of producing realistic fake news videos—created with the same tools others use to design logos or answer math problems.

AI Sometimes Spills Over Its Guardrails

We are in a period of rapid innovation. Things are moving so fast that ordinary users can't keep up. There are new tools launching every day. Some generate graphics, others take photos, make appointments, and diagnose medical conditions. The list is fast heading towards "endless". Organizations across all sectors report building integrations with LLMS into their services or businesses.

Unfortunately, many new AI tools are poorly trained, hastily developed, or limited in capability. Since Meta released its LLaMA models under an open-source license, hobbyists, teenagers, startups, hackers, and artists have jumped on the bandwagon to build their own creations—sometimes quite literally at the kitchen table. On the flip side, some of these models have been adapted into smart learning tool that assist with solving questions, translations, summaries, and AI-powered search. When used thoughtfully, such tools can be incredibly helpful—but not everyone uses them wisely.

The Lowdown on AI

AI analyzes information, processes data, and performs tasks. It can tackle vast datasets and crunch numbers fast. But it doesn't interpret nuances or 'read between the lines'. It does not feel grateful when you address it politely.

AI can provide valuable insights or appear to do so. But it should complement human judgment rather than replace it. When we believe it 'understands' us, we get fooled into thinking that we share the load of our (human) responsibility.

Just remember: when things go wrong, AI can't be held accountable for its actions or biases, but we, humans, can and will be.

Tips for Getting the Best out of AI

The development of AI is moving so fast that we often overlook security measures. As a result, AI systems are surprisingly vulnerable to unconventional cybersecurity threats.

They may make incorrect statements and present them as fact. This flaw is known as "AI hallucination."
AI can exhibit bias, especially when answering leading questions on sensitive matters.
It can be manipulated to write harmful content during "prompt injection attacks." That's when an attacker crafts a message specifically designed to make the AI behave in unintended ways. Prompt injection attacks can cause AIs to produce offensive content.
Data poisoning attacks are particularly hazardous. That's when attackers change or poison the data used to train the AI to answer with inaccuracies, offensive content, or incorrect predictions based on confidential user data.

Nine Tips on How to use AI safely

Keep up to date. There's so much news and hype that it becomes hard to follow, but pick a good news resource and stay in the know. It will help better understand AI systems' capabilities, limitations, and potential risks.

1. Use AI responsibly

No matter what it says in the user agreement or privacy policy, free AI tools will probably use everything you say or ask for training purposes. A few AI tools state that their premium clients' data won't be used for training. Read the user agreements and privacy policy if you need an AI for work. Businesses should protect their operations and have a duty of care towards their customers' data. Are you sure you trust the company? Consider consulting an AI legal expert to protect your best interests.

2. Perform Regular Cyber Hygiene Tasks

Never connect to the internet without encryption, like the one that a reliable VPN offers. One of the most significant benefits of a VPN is an added security layer that will encrypt the data you send to the AI interface. It will stop hackers or snoopers from intercepting your data in transit. It will also ensure you receive the data in an unchanged form that you can trust.
Install an antivirus on your device.
Use a strong password.
Regularly update hardware, software, and apps. Updates fix vulnerabilities that can be exploited by malware. Updates also bring new features, or in the case of AI, it has on several occasions been rolled back after discovering rogue or unwanted behaviour.

3. Don't share Personal Information

Don't share sensitive details. That includes full names, addresses, or phone numbers. Don't share financial or medical information. Some say you can share such information if you trust the platform and know how it handles personal data. However, we have learned from the years-long and ongoing Big Tech data sagas that your data is never truly safe.

4. Understand the Privacy Settings

Review the privacy settings on your AI dashboard. The default setting is usually more beneficial to the tech company than you. Change the settings for more privacy, although anonymity is impossible. Remember that you had to create an account to use the AI tool. That means that you are not a cipher. Your identity is known to the company that created your email account. A tiny bit of digging could reveal your social media accounts — indeed, your entire digital life — to the owner of the AI tool.

5. Be Careful Around Customer Service ChatBots

Customer service chatbots are proliferating across the internet. It's fast becoming your first point of contact with the customer service or sales department. But where did the bot come from, and who trained it? Did the company spend hundreds of thousands or more training their private AI chatbot, or is it a generic SaaS chatbot that will share everything you enter with its creator? Do not share personal information or engage in chats that make you uncomfortable.

6. Be a Critical Thinker

Remember that AI systems are not perfect. They make mistakes. Check facts and cross-reference information before reaching conclusions. Don't make decisions based solely on AI-generated output.

7. Report Inappropriate Content

Report it if you encounter offensive or harmful content or experience inappropriate interactions.

8. It Bears Repeating That AI Is Not Human

AI has no emotions. It cannot understand human frailties or quirks. It cannot replace human relationships. Use it responsibly and always double-check its replies and sources. If possible, do an entry-level online course to get more comfortable with the basics. See AI no more than the clever tool that should free us from routine and allow our creativity to blossom.