How to use ChatGPT and still be a good person

Just a couple of years ago, virtual assistants like Siri and Alexa were punchlines for dinner-party jokes. Now, the tables have turned entirely.

Within weeks of launching, ChatGPT proved it could draft convincing business pitches, write code, and even help people win programming contests. Meanwhile, Lensa AI exploded on social media by transforming mundane selfies into stylized digital avatars. But right behind the viral hype came a wave of messy, unintended consequences.

The New York Times published an ethical gut-check on how to navigate this new wave of generative tools without causing harm.

Teachers are already catching students turning in essays generated by bots. Some women who fed normal portraits into Lensa got back sexualized avatars making them look skinnier, younger, or practically unclothed. As computer scientist Brian Christian put it, we are at the start of a massive societal shift, and the immediate headache for schools is simple: what does homework even mean anymore?

A fancy predictive keyboard that can't fact-check itself

To use these tools responsibly, you have to understand the trick behind the curtain.

Strip away the conversational polish, and ChatGPT is essentially a superpowered version of your smartphone's predictive text. It digests colossal piles of web scrapings to guess which word should come next. Crucially, it has no built-in mechanism to verify whether what it spits out is actually true.

Melanie Mitchell, an AI researcher at the Santa Fe Institute, warned that these bots fool us into thinking they understand far more than they do. If you ask ChatGPT to analyze a classic novel and the web contains faulty interpretations, the bot weaves those mistakes into a confident, polished essay. Share that text, and you are actively spreading misinformation.

Coding carries the same hidden debt. Bot-generated code mimics past solutions, meaning it regularly pulls in outdated patterns and unpatched security vulnerabilities. When asked about these blind spots, OpenAI declined to comment.

The mirror problem and the privacy cost

Image generation carries an equally fraught baggage around bias, ownership, and privacy.

Lensa developer Prisma Labs argued that the hyper-sexualized renders of women simply reflect web data, claiming the AI is merely holding up a mirror to society. But that mirror amplifies harmful tropes without warning, all while borrowing heavily from independent artists who get zero credit or compensation.

Then there is personal privacy. Prisma Labs claims it deletes uploaded selfies within 24 hours, yet privacy advocates urge caution. Evan Greer of Fight for the Future pointed out that feeding your files to AI gives developers the exact fuel they need to build their next models, so you should always assume your private data is accessible.

Why it matters

When a tool seems magic, the initial temptation is to let it do the whole job for us. But we think the smart play right now is treating generative AI as a thinking partner that expands your capacity, not a surrogate that replaces your judgment.

For knowledge workers and students, using it to kickstart an outline or debug a syntax error is a genuine leap in productivity. But outsourcing your final output to a bot is reckless, because the human user remains entirely on the hook for the factual errors, security holes, and plagiarism it introduces.

For the AI industry, the honeymoon of scraping the public internet with zero friction is coming to a close. Bias mitigation and data provenance are about to shift from PR talking points to existential legal battlegrounds. The technology is here to stay, but being a responsible user means keeping your own hands firmly on the wheel.