Ethical artificial intelligence (AI) refers to the development and deployment of artificial intelligence systems that adhere to principles of fairness, transparency, accountability, and respect for human rights. It aims to mitigate biases, ensure data privacy, and prioritize the well-being of individuals and society as a whole. By integrating ethical guidelines, AI systems can operate responsibly, balancing innovation with social responsibility to build trust and avoid unintended harm.¹

From the early days of computers, people have feared and fantasized about the future powers of technology and the idea that we could create a computer system smarter than ourselves. If AI continues to advance at its current rate, some once far-fetched hopes (and fears) about the capabilities of future technology may quickly materialize. With tech giants like Mark Zuckerberg being known for phrases like “move fast and break things,” it’s no surprise that some people are fearful that AI technology may develop faster than we can establish ethical guidelines to contain it. 

Although defining ethical artificial intelligence is a field of study with constantly expanding parameters, for simplicity, we’ll summarize the four main goals of AI ethics: 

• Traceability: The companies and developers who work on our technology should be able to explain the software and internal decisions of the algorithms. This is vital to the maintenance of working AI and the protection of our own safety so that when issues arise, the cause can be located and addressed efficiently.²
• Fairness: AI programs are often developed through machine learning, which involves training machines to learn like humans. Unfortunately, human decision-making is influenced by biases and prejudices, and because humans create training programs and datasets, we often program our preexisting biases into them. 
• Responsibility: We must establish parameters around who bears responsibility for the impacts of AI. This means that as AI systems are developed, deployed, and used by real people, the responsibility for any harmful outcomes must be assigned to liable parties and they should be held accountable.
• Security: Security and privacy are foundational to the responsible development and use of AI systems, particularly as they often run on the sensitive data of real people. This isn’t just a technical necessity but also a societal obligation so that AI systems operate safely, ethically, and legally. 

In short, ethical AI refers to the design, development, and distribution of artificial intelligence systems that prioritize traceability, fairness, responsibility, and security, respecting human rights and societal values.¹

Machine Learning (ML): The training process used to create the algorithms upon which artificial intelligence systems operate. This involves computers making predictions and learning from labeled or unlabeled data that is supplied to it by a human—but no explicit programming of instructions is required. 

Algorithms: Step-by-step procedures or formulas used to perform computations, process data, and solve problems. In the context of AI, algorithms are the core mechanisms driving the learning and decision-making processes based on input data and predefined rules.³  

Turing Test: Developed in 1950 by Alan Turing, this test is a method for determining if a machine can think like a human. To conduct the test, an interrogator asks questions to the computer and a human in a specific subject area, using a set format and context. If the interrogator can't tell which entity is human and which is a computer, the computer passes the test. This test has been vital in studying how machines interact with humans, and for defining "thinking" and "intelligence."³ 

​​Generative Artificial Intelligence (GenAI): A specific type of artificial intelligence that involves the creation of new content based on a prompt or instruction provided by the user. This is achieved through algorithms that attempt to mimic human intelligence and response patterns. Outputs include text, sound, code, images, or video—with the repertoire steadily expanding. For reference, ChatGPT, DALL-E, and Sora are all examples of GenAI tools developed by OpenAI.³ 

Large Language Models (LLMs): A type of GenAI concentrated on all things text and language processing. LLMs can understand vast amounts of data and return human-like responses. When prompted correctly, they perform various tasks ranging from writing code for developers to answering customers’ questions as chatbots.² OpenAI’s ChatGPT and Google’s Gemini are both considered to be LLMs.³ 

Big Data: Large and complex datasets that are difficult to manage and analyze with traditional tools like spreadsheets. AI is often trained on big data, which can include structured data, unstructured data, and mixed datasets, originating from a variety of sources including social media, weather satellites, and smart devices.³ 

Data Ethicist: A scientist who studies the societal effect of technology and data and provides recommendations for other data professionals. Data ethicists must consider fairness, accountability, laws, moral dilemmas, and risks in the creation of potential future technologies, data products, and policies.³  

The advent of new technology always begs the question of how it will eventually be utilized.  Even in the early days of computer engineering, researchers at the forefront of the technology recognized a need for ethical oversight, especially in the wake of World War II. While working on the automation of cannons, computer engineer and professor Norbert Wiener wrote about computers:⁴

“… we are already in a position to construct artificial machines of almost any degree of elaborateness of performance. Long before Nagasaki and the public awareness of the atomic bomb, it had occurred to me that we were here in the presence of another social potentiality of unheard-of importance for good and for evil.”

The earliest foundation of what we would recognize today as “artificial intelligence” ​​dates back to the 1950s. Computer scientist Alan Turing's work on machine intelligence introduced the Turing Test, which sparked debates about the limitations of human intelligence and basic ethical considerations as early AI pioneers focused on creating intelligent systems with little attention to societal implications. In the 1960s, computer scientist Joseph Weizenbaum introduced ELIZA, an early “chatbot” that could take on the persona of a psychiatrist, raising questions about human interaction with machines and the ethical boundaries of AI in psychological contexts in particular. It was around this time that the first warnings emerged about the potential societal impact of AI, including risks of automation and job displacement.⁵ At that time, economists even came close to implementing a Universal Basic Income (UBI) in the United States to protect against the technological disruptions that may have been on the horizon due to AI. 

By the 1980s, as access to AI systems expanded, it became clear that many of them had issues with replicating human biases in their decision-making. One of the first major studies to document this problem was at a medical school in London. The vice dean at the university had the idea to leverage AI algorithms to automate some of the admissions process as a way to make it more efficient and objective. Unfortunately, the plan backfired; instead of making the admissions selections more fair, the algorithm contributed to more discrimination. As the number of women and people of color admitted to the program dwindled, investigators looked into the coding of the algorithm itself. They found that applicants with non-Caucasian names automatically had 15 points taken off their score, and women’s names were marked down by an average of three points. Although the system wasn’t programmed to do this directly, because the algorithm had learned from years of the discriminatory practices of the human admissions officers before it, the same biases had been cemented into the system.6 This public case in particular prompted necessary research on the importance of transparency and accountability in AI decision-making.

With the expansion of the internet in the 1990s, AI’s role in data processing and analytics continued to grow. As developers gained access to more and more data, concerns over data privacy and surveillance ethics began to emerge. Unfortunately, the momentum of the tech industry has always outpaced the speed of regulatory oversight, and the powerful tech titans’ pursuit of profit has largely overshadowed concerns for caution. The role of machine learning and big data continued to fuel AI advancements into the 2000s, amplifying ethical dilemmas like algorithmic bias and discrimination. As public awareness of these issues grew, some organizations began forming ethics committees, but comprehensive oversight remained limited.⁵

In the 2010s, ethical AI discussions went mainstream, largely prompted by high-profile AI failures. One such failure involved biased facial recognition systems gaining public and academic scrutiny. In a 2018 study, three facial recognition programs were analyzed. Researchers found that the average rate of error in determining the gender of fair-skinned men was around 0.8%. However, in darker-skinned women, the error rates skyrocketed to between 20-34%. Why did this happen? It was likely because the data sets used to train the programs were 77% male and 83% white.⁷ This brought public attention to the need for diversity in all aspects of AI development, and AI’s propensity to amplify discrimination. Writers like Cathy O’Neil exposed further issues with AI, particularly in its applications to real-world uses like hiring practices, finding that AI still showed significantly biased decision-making. It was around this time that industry-led initiatives like the Partnership on AI and other academic research centers like the MIT Media Lab started addressing ethical AI. Some governments and global organizations, such as the EU, also introduced AI ethics guidelines emphasizing fairness, accountability, and transparency. These types of broad oversight and regulation committees, although often underdeveloped relative to the complexity of the technology they’re supposed to regulate, have normalized the need for organizations devoted specifically to managing AI-related development. 

In the last few years, the conversation around ethical AI and regulation has been brought to a global stage. Countries have adopted frameworks like the EU's AI Act, emphasizing legal oversight and ethical standards. Many major tech companies have established dedicated teams to audit AI for fairness and reduce bias, though many challenges persist.⁵ With the popularity and accessibility of generative AI products like ChatGPT, the ongoing debates about the ethical use of AI have only expanded. Instead of discussions about AI use taking place only among professors, government officials, or tech leaders, AI and its appropriate usage has become a non-theoretical, pressing issue for many ordinary people. Questions of copyright, misinformation, education, user safety, and many more unexpected dilemmas have quickly become a frequent topic of discussion. All of this has further emphasized the need for robust ethical frameworks. The question is: will AI’s development outpace experts’ ability to control it? 

Norbert Wiener

As a Massachusetts Institute of Technology professor and pioneer of cybernetics, Wiener foresaw the potential ethical challenges posed by intelligent systems. In his writing, he warned about the societal impact of automation and the importance of aligning technological systems with human values, laying the groundwork for later debates on AI's role in society and ethics.⁴

Joseph Weizenbaum 

A computer scientist known for creating ELIZA, an early natural language processing program that controversially simulated a conversation with a therapist. Weizenbaum became a prominent critic of AI, arguing that humans shouldn’t delegate decisions to machines, particularly in morally sensitive domains like healthcare or warfare.⁵

Alan Turing

Born in 1912, Turing was an English mathematician, computer scientist, logician, cryptanalyst, philosopher, and theoretical biologist. He developed the Turing test, which aimed to test whether computers could ‘pass’ as humans, and sparked widespread conversations about the definition of intelligence.⁵

Cathy O'Neil 

Data scientist and author of the book Weapons of Math Destruction (2016), O’Neil’s work exposes how algorithms can perpetuate bias and inequality. Her writing has highlighted the dangers of AI systems that lack accountability, particularly in areas like hiring, policing, and credit scoring.⁸

Ensuring ethical AI development and use is important for users on both a personal level and a societal level. Although the ethics of AI have broad impacts, we’ll focus on the four main areas of traceability, fairness, responsibility, and security. 

Traceability 

As AI continues to evolve in complexity, it’s increasingly important that we ensure our AI models are explainable and that their decision-making processes are transparent. From a consumer perspective, as AI is integrated into more aspects of our daily lives, users should be certain they can rely on their technology and aren’t surprised by the outcomes or behaviors of their tools. The companies and developers who work on the technology should be able to explain the software and all of its internal decisions so that when issues arise, the cause can be identified and addressed efficiently (so we don’t end up with out-of-control robots like in the horror movie M3GAN, about a rogue AI doll). Perhaps it’s not a coincidence that so many scary movies revolve around the premise of advanced AI systems taking on a will of their own, becoming unstoppable, and trying to outsmart and overpower humans (Ex Machina, The Terminator, Her, and Westworld, just to name a few). 

Fairness

Machine learning algorithms have consistently come under scrutiny for being discriminatory or inaccurate towards racial or gender minorities. For example, facial recognition technology often performs far worse for subjects who are female and Black, likely because the benchmark dataset for facial recognition skews 70% male and 80% white. In order to create more ethical software, we need to include more diverse data in our ML datasets when we train AI models, using data that actually reflects the diverse population the machines are meant to serve. Because it’s developed and used by imperfect, biased, humans, AI will often reinforce the same biased or discriminatory patterns that humans have.⁹ We must work to design ethical AI that’s not only free of bias but could even help us identify and combat errors in our own decision-making. 

Responsibility 

As AI becomes increasingly independent, we need to establish parameters around who bears responsibility for its impacts. A common thought experiment used to illustrate the issue of assigning responsibility for AI involves a self-driving car: if the car crashes, who is to blame? The person who owns the car? The engineer who programmed the software? The company that designed the car? Can we hold a machine accountable? Any major organization, including our governments, needs to ensure that there is an established chain of responsibility prior to something going wrong. 

Unless accountability is clearly established, developers may be overly risk-averse. In the case of self-driving cars, these types of vehicles, although imperfect, are often involved in fewer fatal accidents than human drivers. However, if drivers are worried they will take the blame for any accidents in their automated vehicles, or developers are concerned they will be held liable, then both parties may decide it’s not worth the personal risk of adopting or continuing to expand the new technology, even though it’s likely to save more lives overall. On the other hand, if reasonable actions aren’t taken to hold people responsible for the damaging (or even fatal) consequences of AI, then developers and users alike may be more reckless with the technology, designing new algorithms or deploying new features without adequate testing, or using the products in risky situations. Currently, the law states that the liability for any problems lies with whoever caused the damage or might have foreseen the product being used in the way that caused the issue. Of course, this still leaves ample room for debate, and as new cases emerge, the issue of responsibility will continue to evolve. 

Security

It sometimes seems like each new generation of smart devices and every social media update collects even more personal data than the previous version of the technology. As the data collected on us increases, so does the need for data security. Most AI systems rely on big data—such as personal, financial, or medical data—and this information is often used to train AI algorithms to guide how they function. This sensitive information needs to be protected, and data ethicists often argue over who should have access to it. Particularly in situations where users don’t know or can’t control who has access to their information, who should be allowed to see users’ personal details? What kind of information can or should be bought and sold by companies looking to profit? While there’s still debate over exactly what data privacy can or should mean, almost everyone can agree that there is a desperate need for data security to prevent outright malicious or illegal activity. 

Because AI programs have access to so much of our personal data, insufficient data privacy creates opportunities for unauthorized access or misuse, from identity theft to discrimination to government surveillance. As we weave AI systems into more of our critical infrastructures, the risks associated with cyberattacks have only increased, as evidenced in recent incidents threatening diverse sectors from healthcare in the 2024 Change Healthcare ransomware attack¹⁰ to finance in the 2023 Bank of America data breach.11 Whether because of intentional exploitation or catastrophic oversight, a lack of adequate data security can leave people vulnerable or disrupt vital systems. In order for AI to gain wider acceptance, users have to trust that their information is secure, their privacy is respected, and the systems are reliable. If security breaches or unethical handling of data become more common, the public will likely lose confidence in AI systems and the organizations behind them.

Agreeing on ethical design and use of any new technology is tricky; no criteria will ever be perfect, and there will always be disagreement. Let’s consider some of the biggest conversations about ethical use of AI:

Environmental Responsibility 

We’ve discussed general responsibility as a tenet of ethical AI. But when we question the chain of command and who takes ultimate ownership of AI systems’ ramifications, we must also ask: who is responsible for the impact on the planet?  

While scientists are actively working with AI to develop solutions to our climate crisis, the energy usage and electronic waste caused by AI technology are paradoxically speeding up the process of climate change. The microchips that power AI often require rare earth metals, which rely on inhumane and environmentally destructive mining practices. Most large-scale AI companies depend on cloud service providers, which are often comprised of massive data centers that consume and dispose of an unfathomable amount of raw materials. In fact, making a two-kilogram computer requires about 800 kilograms of raw materials.¹² The data centers and their electronic waste often contribute to the pollution of nearby areas with dangerous substances like lead and mercury. The centers also rely on large amounts of water, both during construction and for cooling the many electrical components once operational. According to one estimate, AI-related infrastructure could soon consume six times the annual water usage of Denmark. As the number of AI programs and devices has boomed, so has the number of data centers required to store all their data. Just over a decade ago, there were roughly 500,000 of these data centers, but there are now more than 8 million.¹² 

All of these data centers, as well as the personal devices and businesses relying on AI technology, consume far more energy than we likely realize. One ChatGPT search requires as much energy as it would take to power one 5-watt light bulb for 20 minutes. Now, take a minute to think about how many ChatGPT inquiries that one coworker of yours has run in the last week alone. That’s a lot of electricity. As AI’s popularity continues to skyrocket, so will our energy consumption. Organizations like the International Energy Agency are fearful of these impacts; in a small but tech-heavy country like Ireland, for example, data centers are expected to account for nearly 35% of the country’s energy usage by 2026.¹² 

Political Influence 

Already, AI has been at the center of political scandals. The most infamous is perhaps the Cambridge Analytica data scandal, where it was revealed that the politically conservative consulting firm had illegally harvested the data of millions of Facebook users and used AI-driven targeting to influence the 2016 U.S. Presidential election and Brexit referendum.¹³ Instances of data misuse have undermined public trust in political institutions and may play a role in the public’s increasing vulnerability to misinformation. As AI’s capabilities increase, so will the ease of spreading misinformation. Already, AI-generated deepfakes are being used to alter perceptions of political figures and manipulate electoral outcomes. AI-driven social media bots amplify fake news and create echo chambers that reinforce existing biases and polarization. With great power comes great responsibility, and developers of ethical AI will need to find bipartisan ways to moderate attempts to use AI for unethical political influence. 

AI in Education 

If you’ve been a student in the last few years, or if you know any students or education professionals, you’ve likely discussed or at least considered the implications of AI in education. Often, the conversations in classrooms revolve around the importance of academic integrity, with teachers reminding students that their work must be their own, and that they can’t, for example, just ask ChatGPT to write their essay on The Great Gatsby for them. While some educators have doubled down on banning AI technology from the classroom and forbade their use for homework assignments, others have acknowledged that this technology is not going away, with some teachers trying to integrate AI into their syllabi, teaching students how to learn with its support. While educators struggle with how to fairly grade AI-assisted work, many students grapple with the ethics of using AI in situations like the high-stakes college admissions process, where they may be at a disadvantage if they’re in the minority of students not using this asset. 

In addition to the ethical dilemmas students and teachers face as they navigate new definitions of plagiarism, critical thinking, and academic integrity, there are many other concerns as to how we can create and implement ethical AI in education. While AI tools like generative tutoring systems can enhance learning by personalizing instruction, their widespread adoption risks deepening the digital divide. As these types of tutoring or homework assistance programs become the norm, students without access to reliable internet or technology may be left behind, worsening existing inequalities. As these same AI tutoring systems gain traction, the traditional roles of teachers and tutors may be called into question. If teachers and tutors are replaced, questions of responsibility arise if and when the systems malfunction or fail to help students. It’s easy to imagine other ethical AI failures in schools, perhaps exacerbating preexisting discrimination by recommending certain education pathways to some students and not others or through unequal grading practices. Particularly with children, data privacy and security are vital, and creating ethical AI that is safe and fair is essential. 

Facebook’s “Emotional Contagion” Study 

In 2014, Facebook made headlines, not for a fun new AI chatbot or game, but because they were part of a controversial study published in a prestigious science journal. In the experiment, researchers spent a week altering the content presented in the news feeds of almost 700,000 Facebook users. The goal was to assess whether or not what people post is influenced by exposure to emotional content by one’s contacts. There is a popular narrative that seeing constant posts of your friends’ happy moments can make people depressed— and while this could still be true internally, the study found that exposing users to more positive messages coming from their friends actually prompted people to share more of their own positive content.¹⁴ Today, AI algorithms play a huge role in shaping the content we’re exposed to.

The bigger reason this story gained so much attention, though, was because of the public backlash when Facebook users discovered they had unknowingly been experimented on. People were outraged at the idea of researchers manipulating their emotions through Facebook. The “emotional contagion” study was suddenly being critiqued by journalists and scholars around the world, who questioned the ethics of the study, asking: should researchers (or AI algorithms) have the right to alter what content people see on commercial sites without ethics oversight or informed consent? One might ask: what is informed consent on the internet in the first place? Although we often agree to certain terms and conditions when we sign up for new websites or technology, when was the last time you actually read the agreement in full? Is it fair to hold people responsible for something they’ve technically signed, but we have evidence to suggest that they’ve not actually read (through metrics like tracking time on each webpage screen)? 

Even if we decided that it is ethically permissible to manipulate users’ feed content (and possibly their emotions in the process), who is allowed to make these adjustments? Some critics were particularly concerned with the source of the study: instead of the research being strictly academic, the manipulation was partially guided by a profit-driven business. While many social scientists and ethicists argued over the ethics of the study, many computer scientists and industry practitioners came forward in its defense. They argued that A/B testing is a central part of designing successful AI algorithms, which run tools that give personalized recommendations and curate Facebook’s news feed. So, why do people perceive the same practice differently when it’s part of academic research compared to when it’s undertaken as business practice? The answer is that academic research is typically held to a higher ethical standard than business practices and, therefore, is assessed using different criteria. As more businesses make use of AI algorithms and big data, these questions will continue to impact a wider population. Data ethicists, governments, and industry leaders will need to evaluate what is considered acceptable AI business and research practice, and if it’s dependent on who is conducting the research.

Air Canada Chatbot Misinformation 

In February 2024, after the death of his grandmother, a man named Jake Moffatt used Air Canada’s virtual assistant to inquire about bereavement fares. The chatbot advised him to buy his roundtrip plane tickets from Vancouver to Toronto at the regular price and informed him that he could apply for a bereavement discount within 90 days of purchase. After spending almost CA$1700, Moffatt submitted his refund claim, and the airline turned him down, saying that bereavement fares can’t be claimed after tickets have been purchased. Clearly, the chatbot had given Moffatt false information. When he took Air Canada to court, claiming that the airline was negligent and the chatbot misrepresented important information, the airline argued that it couldn’t be held liable for the information provided by its virtual assistant. Ultimately, Moffatt and his lawyer won the case and were refunded, as the airline didn’t take “reasonable care to ensure its chatbot was accurate.”¹⁵ Cases like this underscore the question of accountability: when AI goes awry, who should be held responsible? In this situation, the courts ultimately decided that the airline was at fault, but within the organization, who should take the blame? If they had used a third-party chatbot developer, would the algorithm designers share the responsibility? As AI proliferates into more and more fields, theoretical questions of responsibility are increasingly raised in real-world situations. 

Tay’s Racist Tweets

You may remember the disaster that occurred in March 2016 when Microsoft attempted to use social media interactions as training data for machine learning algorithms. The company introduced Tay, an AI chatbot designed as an experiment in “conversational understanding,” to Twitter. Modeled after the persona of a teenage girl and preloaded with anonymized public data and content crafted by professional comedians, Tay used a combination of machine learning and natural language processing technology to interact with Twitter users. Unfortunately, instead of evolving into a more sophisticated, human-like chatbot, within 16 hours, Tay posted over 95,000 tweets, many of which were overtly racist, misogynistic, and antisemitic.¹⁶ That’s perhaps because Microsoft had forgotten one important component: the uncontrolled variable of the humans it would be interacting with. Not only was the former Twitter (currently X) filled with profanity, but once people learned about the bot, they began tweeting hateful messages to it, and, as Tay was designed to do, the bot learned from these messages and mimicked accordingly. Although Microsoft quickly suspended and ultimately discontinued the chatbot, its impact sparked a wider awareness of the many potential pitfalls that can come with AI trained on large sets of public data— and the risks of releasing technologies before they’ve been thoroughly tested. 

“It’s Complicated:” An Ode to Our Relationship with AI 

Regardless of how you perceive your impact on AI’s development, you have the power as a user to choose how you will interact with the technology, ethically or otherwise. This piece dives into the practical uses and limitations of the current AI landscape, from which many ethical dilemmas arise. 

Machine Learning 

Understanding the basics of how AI works is important not just for programmers but also for data ethicists, decision-makers, and users. Read here to learn more about what machine learning is, how it has developed over time, and its potential pitfalls. 

1. Khan, A. A., Badshah, S., Liang, P., Waseem, M., Khan, B., Ahmad, A., Fahmideh, M., Niazi, M., & Akbar, M. A. (2022). Ethics of AI: A systematic literature review of principles and challenges. EASE '22: Proceedings of the 26th International Conference on Evaluation and Assessment in Software Engineering, 383–392. https://doi.org/10.1145/3530019.3531329 
2. Kieslich, K., Keller, B., & Starke, C. (2022). Artificial intelligence ethics by design. Evaluating public perception on the importance of ethical design principles of artificial intelligence. Big Data & Society, 9(1). https://doi-org.gate3.library.lse.ac.uk/10.1177/20539517221092956 
3. Krpan, Dario. (2023). Behavioural Science in An Age of New Technology. Lecture 1,2,10: PB434
4. Wiener, N. (1948). Cybernetics: Or control and communication in the animal and the machine. Technology Press. 
5. Borenstein, J., Grodzinsky, F. S., Howard, A., Miller, K. W., & Wolf, M. J. (2021). AI ethics: A long history and a recent burst of attention. Computer, 54(1), 96–102. https://doi.org/10.1109/MC.2020.3034950 
6. Schwartz, O. (2019, April 15). Untold history of AI: Algorithmic bias was born in the 1980s. IEEE Spectrum. https://spectrum.ieee.org/untold-history-of-ai-the-birth-of-machine-bias 
7. Buolamwini, J., & Gebru, T. (2018). Gender shades: Intersectional accuracy disparities in commercial gender classification. In S. A. Friedler & C. Wilson (Eds.), Proceedings of Machine Learning Research: Conference on Fairness, Accountability, and Transparency (Vol. 81, pp. 1–15). MIT Media Lab. 
8. O'Neil, C. (2016). Weapons of math destruction: How big data increases inequality and threatens democracy. Crown Publishing Group. 
9. Nouri, S. (2021, February 3). Council post: The role of bias in artificial intelligence. Forbes. https://www.forbes.com/sites/forbestechcouncil/2021/02/04/the-role-of-bias-in-artificial-intelligence/?sh=1751699e579
10. Alder, S. (2024, November 19). Change Healthcare fully restores clearinghouse services after February ransomware attack. HIPAA Journal. 
11. Winder, D. (2024, February 13). Bank of America warns customers of data leak following 2023 hack. Forbes. https://www.forbes.com 
12. United Nations Environment Programme. (2024, September 21). AI has an environmental problem. Here’s what the world can do about that. UN Environment Programme. https://www.unep.org 
13. Hinds J, Williams EJ, Joinson AN (2020) “It wouldn’t happen to me”: Privacy concerns and perspectives following the Cambridge analytica scandal. International Journal of Human–Computer Studies 143: 102498. 
14. Kramer, A. D. I., Guillory, J. E., & Hancock, J. T. (2014). Experimental evidence of massive-scale emotional contagion through social networks. Proceedings of the National Academy of Sciences of the United States of America, 111(24), 8788–8790. https://doi.org/10.1073/pnas.1320040111 
15. Yagoda, M. (2024, February 23). Airline held liable for its chatbot giving passenger bad advice - what this means for travellers. BBC News. https://www.bbc.com/travel/article/20240222-air-canada-chatbot-misinformation-what-travellers-should-know
16. Victor, D. (2016, March 24). Microsoft created a twitter bot to learn from users. it quickly became a racist jerk. The New York Times. Retrieved 2024, from https://www.nytimes.com/2016/03/25/technology/microsoft-created-a-twitter-bot-to-learn-from-users-it-quickly-became-a-racist-jerk.html.