The short answer on artificial intelligence pros and cons: AI tools can save real time, speed up science and help people with disabilities. But they also make confident mistakes, can repeat bias, and run on data centers that use a growing share of electricity and water. Both sides are backed by data, so the useful question is where AI helps, where it fails, and how to use it with care.
This guide sets out the evidence on each side as of October 2026, with sources, so you can judge for yourself. For more explainers like this one, see our science and technology guides.
Why artificial intelligence is important: the benefits
Productivity at work
The best evidence comes from studies of real workers, not from company claims. One well-known study followed 5,179 customer support agents at a US software firm. Agents with an AI assistant resolved about 14% more issues per hour, according to the working paper. The published version in the Quarterly Journal of Economics (2025) puts the average gain at about 15%. Newer and less experienced agents gained the most, because the tool passed on the habits of top performers.
Still, the gains are not automatic. In a 2025 randomized trial by the research group METR, 16 experienced open-source developers took about 19% longer on tasks when AI tools were allowed. Yet they believed they had been faster. So results depend on the task, the tool and the person using it.
Science and medicine
AI has already changed some parts of research. The 2024 Nobel Prize in Chemistry went in part to the creators of an AI model that predicts the 3D shape of proteins. That problem had puzzled biologists for about 50 years, and the predictions now help with drug and enzyme research.
In health care, the FDA keeps a public list of AI-enabled medical devices it has authorized. It holds well over 1,000 entries, with new ones added as recently as mid-2026. Most are in radiology, for example software that flags possible problems on scans for a doctor to review. The doctor still makes the call.
Accessibility
For many people, the most practical benefit is access. Live captions help deaf and hard-of-hearing users follow calls and videos. Image description tools read signs, menus and mail aloud for blind users. Speech-to-text helps people who cannot type easily. Translation tools also lower language barriers for immigrants dealing with schools, doctors and landlords.
Why artificial intelligence is bad: the main risks
Errors and “hallucinations”
Chatbots can produce false statements in a fluent, confident tone. The National Institute of Standards and Technology calls this “confabulation.” It lists it among 12 risks in its Generative AI Profile (NIST AI 600-1), published in July 2024. Made-up court cases, wrong drug doses and fake quotes have all turned up in real use. That is why anything with legal, medical or money consequences needs a human check.
Bias
Models learn from data created by people, so they can pick up the same unfair patterns. If past hiring or lending data favored some groups, a model trained on it may do the same. In practice, the risk is highest when AI scores people for jobs, housing, credit or insurance.
Privacy and security
Anything you type into an online AI tool may be stored. Depending on the settings, it may also be reviewed or used to train future models. AI also helps criminals. The Federal Trade Commission has warned that scammers can clone a relative’s voice from a short clip posted online and use it in fake emergency calls. Phishing emails have also become harder to spot, because the grammar mistakes that once gave them away are gone.
Misinformation
Fake images, audio and video are now cheap to make. That makes it easier to spread false stories. It also lets people dismiss real recordings as fake. Labels and watermarks help, but they are not yet reliable across all platforms.
AI and jobs: what the data says
Job fears are common, so the numbers matter. In January 2024, the International Monetary Fund estimated that almost 40% of jobs worldwide are exposed to AI, and about 60% in advanced economies. Exposure is not the same as job loss, though. The IMF expected roughly half of exposed jobs to benefit from AI, while the other half could see lower demand and wages.
US projections are more specific. In a July 2026 brief on the 2024-34 employment outlook, the Bureau of Labor Statistics expects AI-driven efficiency to limit some office roles. For example, it projects customer service representatives to fall 5.5%, or about 153,700 jobs. Legal secretaries and administrative assistants are projected to fall 5.8%. At the same time, it projects data scientists to grow 33.5% and software developers 15.8%. In short, the change looks gradual and uneven, not sudden.
Is artificial intelligence bad for the environment?
AI itself is software, but it runs in data centers that need power and cooling. The International Energy Agency estimates that data centers used about 415 terawatt-hours of electricity in 2024, roughly 1.5% of world demand. In its base case, that figure more than doubles to about 945 TWh by 2030, with AI as the main driver.
The US is the largest user. A 2024 report by Lawrence Berkeley National Laboratory for the Department of Energy found that data centers used about 4.4% of US electricity in 2023 (176 TWh). It projects 6.7% to 12% by 2028. The same report estimates direct cooling water use of about 17 billion gallons in 2023. Far more water is used indirectly to generate the power.
Whether this is “bad” depends on where the power comes from and where the centers are built. A data center on a clean grid in a wet region has a smaller footprint than one on fossil power in a dry area. Efficiency also keeps improving. However, total demand is rising faster than efficiency gains, which is why utilities and regulators now watch it closely.
AI pros and cons at a glance
| Area | Benefit | Risk |
|---|---|---|
| Work | About 14-15% more output in a large customer support study | Some experts slowed down; some office jobs projected to shrink |
| Science | Protein structure prediction won a 2024 Nobel Prize | Errors can slip into research if results are not checked |
| Health | FDA-authorized tools help doctors read scans | Bias in training data can hurt some patient groups |
| Accessibility | Captions, image descriptions, speech-to-text | Tools can still mishear or misdescribe |
| Information | Fast summaries and translation | Hallucinations, deepfakes, misinformation |
| Privacy and security | AI helps spot fraud and malware | Data storage, voice-cloning scams, better phishing |
| Environment | Chips and cooling keep getting more efficient | Data center power use could double worldwide by 2030 |
How AI regulation is developing in the US and California
As of October 2026, there is no comprehensive federal AI law. Instead, rules come from existing laws on fraud, discrimination and privacy, plus a growing set of state laws.
California has moved furthest. Governor Newsom signed SB 53, the Transparency in Frontier Artificial Intelligence Act, on September 29, 2025. It took effect January 1, 2026. The law covers developers of the largest models, those trained with more than 10^26 computing operations. Companies with over $500 million in yearly revenue must publish a safety framework. Developers must also report critical safety incidents to the state Office of Emergency Services. Employees who raise safety concerns get whistleblower protection, and fines can reach $1 million per violation.
Two other California laws also took effect January 1, 2026. SB 243 sets rules for “companion” chatbots, such as reminding users they are talking to a bot. AB 2013 requires developers of generative AI to post a summary of their training data.
At the federal level, an executive order signed December 11, 2025 (EO 14365) directs the Justice Department to challenge state AI laws it sees as too burdensome. In March 2026, the White House also asked Congress to pass a single national standard. But an executive order cannot by itself override state law, and Congress has not passed a preemption bill. So state laws such as SB 53 remain in effect for now. This is general information, not legal advice.
Chatbots have come a long way since 2018, when firms were still partnering just to measure how enterprise chatbots performed. Back then, the main worry was whether bots understood simple requests. Today, lawmakers focus on safety reports, deepfakes and child protection.
How to use AI tools safely
- Check facts that matter. Treat AI answers as a first draft. Verify names, numbers, laws and medical advice against an official source.
- Keep private data out. Do not paste Social Security numbers, passwords, medical records or client files into public tools. Also review the privacy settings, and turn off training on your chats if the option exists.
- Agree on a family code word. If a “relative” calls in a panic asking for money, hang up and call them back on a number you know.
- Ask for sources. Then open them. A link that does not exist is a sign the answer was made up.
- Know your workplace rules. Many employers now limit which tools staff may use for company data.
- Be skeptical of shocking media. If a clip seems designed to make you angry, look for the same story from a trusted outlet first.
Read next
Frequently asked questions
What are the main pros and cons of artificial intelligence?
The main pros are faster work on routine tasks, help with scientific research, support for medical imaging and better accessibility. The main cons are factual errors, bias, privacy and scam risks, job shifts in some office roles, and rising energy and water use.
Why is artificial intelligence bad?
AI is not bad by nature, but it can cause harm. It can state false things confidently, repeat unfair patterns from its data and make fraud easier. Those risks grow when people rely on it without checking.
Why is artificial intelligence important?
It can handle large amounts of data quickly. For example, it helped solve protein structure prediction and now supports doctors reviewing scans. It also gives people with disabilities new ways to read, hear and communicate.
Is artificial intelligence bad for the environment?
It adds to energy demand. The IEA expects global data center electricity use to roughly double by 2030, and US data centers could reach up to 12% of national electricity by 2028. The impact depends heavily on how clean the local power grid is.
Have a question about this topic? Ask in the comments below. Our editors read every comment and answer questions here, and we add useful answers to the guide.
Sources
- nber.org/papers/w31161
- metr.org/blog/2025-07-10-early-2025-ai-experienced-os-dev-study
- nobelprize.org/prizes/chemistry/2024/press-release
- fda.gov/medical-devices/artificial-intelligence-enabled-medical-…
- nvlpubs.nist.gov/nistpubs/ai/NIST.AI.600-1.pdf
- consumer.ftc.gov/consumer-alerts/2023/03/scammers-use-ai-enhance-their-fa…
- imf.org/en/Blogs/Articles/2024/01/14/ai-will-transform-the-globa…
- bls.gov/opub/ted/2026/artificial-intelligence-information-techno…
- iea.org/reports/energy-and-ai
- energy.gov/articles/doe-releases-new-report-evaluating-increase-ele…
- eta-publications.lbl.gov/publications/2024-lbnl-data-center-energy-usage-report
- gov.ca.gov/2025/09/29/governor-newsom-signs-sb-53-advancing-califor…
- leginfo.legislature.ca.gov/faces/billNavClient.xhtml
- govinfo.gov/content/pkg/FR-2025-12-16/html/2025-23092.htm