Welcome to The AI Ethics Brief, a bi-weekly publication by the Montreal AI Ethics Institute. Stay informed on the evolving world of AI ethics with key research, insightful reporting, and thoughtful commentary. Learn more at montrealethics.ai/about.

💖 Help us keep The AI Ethics Brief free and accessible for everyone. Consider becoming a paid subscriber on Substack for the price of a ☕ or make a one-time or recurring donation at montrealethics.ai/donate.

Your support sustains our mission of Democratizing AI Ethics Literacy, honours Abhishek Gupta’s legacy, and ensures we can continue serving our community.

For corporate partnerships or larger donations, please contact us at support@montrealethics.ai.

• Mapping the Ethics of Generative AI: A Comprehensive Scoping Review

• Careless Whisper: Speech-to-text Hallucination Harms

• AI Framework for Healthy Built Environments

• Microsoft expects to spend $80 billion on AI-enabled data centers in fiscal 2025 – CNBC

• Meta scrambles to delete its own AI accounts after backlash intensifies – CNN

• Don’t panic: AI can strengthen democracy too – Bulletin of the Atomic Scientists

Happy New Year! As we enter 2025, AI has moved beyond its theoretical promise to become a powerful force shaping society. The past year marked a crucial shift from voluntary ethics guidelines to binding regulations, revealing both AI’s transformative potential and its risks. From healthcare and democracy to AI surveillance and civil liberties, developments in 2024 highlighted an urgent truth: the ethical governance of AI is not just a technical challenge but a fundamental social imperative. In no particular order, here are ten defining stories that shaped AI ethics in 2024 and our take on what to expect in 2025.

• The Story
  The implementation of the EU AI Act marked a significant milestone in AI regulation. It introduced a risk-based categorization of AI systems and mandatory requirements for high-risk applications. Complementing this regulatory advancement, international coordination efforts gained momentum with the inaugural meeting of the International Network of AI Safety Institutes in San Francisco on November 20-21, 2024.

• Why It Matters
  The EU’s approach signifies a transition from voluntary guidelines to enforceable standards, particularly in critical sectors like healthcare, financial services, law enforcement and criminal justice, and critical infrastructure. However, this framework raises concerns about potential innovation barriers and regulatory arbitrage.

• Looking Ahead
  2025 will test whether the EU AI Act can effectively balance innovation with protection while inspiring similar frameworks globally. Key challenges include enforcement mechanisms and international regulatory alignment. The incoming Trump administration is also expected to rescind and replace Biden’s executive order on AI. However, the exact details of Trump’s AI policy remain uncertain — some bipartisan-supported measures may be retained or modified rather than completely eliminated.

• The Story
  Leading AI labs have intensified efforts to ensure AI systems align with human values and safety parameters. As reported in The Wall Street Journal, Anthropic’s Frontier Red Team testing exposed critical vulnerabilities, highlighting the need for rigorous safety evaluations.

• Why It Matters
  This emphasis on safety reflects growing recognition that AI development must prioritize human welfare over rapid deployment. However, proprietary safety measures risk creating knowledge asymmetries between large labs and the broader research community. A report by the Institute for AI Policy and Strategy notes that while companies like Anthropic, Google DeepMind, and OpenAI are conducting technical research into safe AI development, there are potential gaps in areas such as model organisms of misalignment, multi-agent safety, and safety by design.

• Looking Ahead
  Balancing open science with responsible development will be crucial. Expect an increased focus on transparent testing protocols and standardizing the process of red teaming AI systems — robust and repeatable processes that accurately reflect model capabilities, establishing a shared baseline on which different models can be meaningfully compared.

• The Story
  Major elections worldwide faced unprecedented challenges from AI-generated disinformation. Notably, the U.S. imposed sanctions on entities in Iran and Russia for attempting to interfere in the 2024 elections through AI-driven influence campaigns. As Lisa Reppell, a researcher at the International Foundation for Electoral Systems, stated, “A world in which everything is suspect — and so everyone gets to choose what they believe — is also a world that’s really challenging for a flourishing democracy.”

• Why It Matters
  AI-powered misinformation poses significant risks to democratic processes, necessitating new approaches to content verification and digital literacy. The sophistication of AI-generated content, particularly deepfakes, makes it increasingly challenging to identify manipulated media, highlighting the need for advanced detection technologies. In response to the growing threat of AI impersonation, Tennessee enacted the Ensuring Likeness, Voice and Image Security (ELVIS) Act, becoming the first state to protect individuals’ voices and likenesses from unauthorized AI-generated replication.

• Looking Ahead
  There will be an increased focus on developing detection technologies and fostering international cooperation to combat cross-border disinformation campaigns. The global nature of AI-generated disinformation requires collaborative efforts to safeguard the integrity of democratic processes.

• The Story
  Generative AI systems demonstrated concerning patterns of bias, particularly in recruitment and creative industries. Research from the University of Washington revealed that AI tools used in resume screenings favored white-sounding names over those associated with Black individuals, and male-associated names over female ones, perpetuating racial and gender disparities in hiring processes.

• Why It Matters
  These biases threaten to encode existing social inequities into automated systems, disproportionately impacting marginalized communities. In recruitment, biased AI tools can lead to discriminatory hiring practices, undermining diversity and inclusion efforts. In creative industries, AI-generated stereotypes can perpetuate cultural misrepresentations, influencing public perception and reinforcing harmful biases.

• Looking Ahead
  The development of more robust fairness metrics and mandatory bias audits may become standard practice. Organizations are encouraged to establish comprehensive policies and procedures for the use of generative AI, including regular audits to detect and mitigate biases, and the implementation of guidelines to ensure ethical AI deployment.

• The Story
  The integration of AI in healthcare has led to significant advancements in medical diagnostics and treatment planning. However, automated decision-making systems have come under scrutiny for their lack of transparency and accountability. The killing of the UnitedHealthcare CEO in New York City has brought attention to systemic issues in algorithmic healthcare management, particularly concerning the use of AI to deny insurance claims. 2024 also saw concerns emerge about AI mental health applications’ impact on teens, highlighting broader issues of AI safety in healthcare.

• Why It Matters
  The rapid adoption of AI in healthcare necessitates a balance between technological innovation and the preservation of patient rights. In insurance, AI systems can significantly impact patient access to care through automated coverage decisions. The lack of regulatory parity between clinical AI tools and insurance algorithms has created a dangerous oversight gap, allowing financially motivated decisions to override medical judgment. This undermines both patient trust and healthcare equity.

• Looking Ahead
  We may see mounting pressure for comprehensive AI healthcare regulation that encompasses insurance algorithms. Key developments will likely include establishing regulatory frameworks that subject insurance AI to the same scrutiny as clinical AI tools, implementing transparency requirements for coverage decisions, and creating appeal mechanisms that prioritize medical expertise over automated predictions.

• The Story
  In 2024, AI tools became increasingly integrated into educational settings, offering personalized learning experiences for students and automating administrative tasks for educators. Platforms like Khan Academy’s Khanmigo provide AI-powered tutoring and teaching assistance, enhancing accessibility to quality education. As Joseph Fuller, a professor at Harvard Business School, noted, “AI literacy is the modern equivalent of typing in the 1970s and ’80s, a universal requirement for all students going into all fields of work.”

• Why It Matters
  The dual impact of AI in education—improving access while challenging integrity—highlights the need for balanced integration strategies. The ease of generating AI-assisted content has led to challenges in distinguishing between original student work and AI-generated material, raising concerns about academic integrity. Educational institutions responded by implementing AI detection tools and revising academic policies to address these issues.

• Looking Ahead
  As AI continues to evolve, the focus must shift from merely developing comprehensive guidelines and implementing robust policies to maintain academic integrity, to equipping educators with the skills and training needed to effectively integrate AI into their teaching practices. Professional development initiatives will play a critical role in helping educators navigate the opportunities and challenges posed by AI.

• The Story
  The expansion of AI-driven surveillance technologies intensified debates over privacy rights and civil liberties. Governments and law enforcement agencies worldwide increasingly adopted AI tools for monitoring public spaces, aiming to enhance security and public safety. France extended its AI-powered video surveillance beyond the 2024 Olympics, raising concerns among privacy advocates about potential overreach and the erosion of civil liberties.

• Why It Matters
  The proliferation of AI surveillance poses significant risks to democratic freedoms. While these tools are often implemented to enhance security, they can inadvertently lead to mass surveillance and abuse if such data falls into the wrong hands or is used without proper safeguards. The American Civil Liberties Union (ACLU) has highlighted the dangers of AI systems that perpetuate discriminatory outcomes and enhance surveillance capabilities, calling for concrete actions to protect civil rights and civil liberties.

• Looking Ahead
  Ongoing vigilance and advocacy from civil society groups, lawmakers, and regulators will be essential to ensuring that advancements in AI do not compromise fundamental human rights. Legislative measures should balance security needs with robust regulatory frameworks that clearly define the permissible use of AI by governments to protect individual freedoms.

• The Story
  UN Secretary-General António Guterres reaffirmed the urgent need to preserve human control over the use of force and to address the harmful effects of lethal autonomous weapons systems (AWS). Released on August 6, a United Nations report compiled input from 73 states and 33 civil society organizations, including Human Rights Watch and the Arms Control Association, highlighting the ethical dilemmas and risks of unintended escalations posed by AWS.

• Why It Matters
  Autonomous weapons systems, often referred to as “Killer Robots,” raise fundamental questions about human control, accountability, and the ethics of delegating life-and-death decisions to algorithms in combat situations. Without robust oversight, the deployment of such systems risks violating international humanitarian law, escalating conflicts unintentionally, and resulting in the loss of innocent lives. Clear legal and ethical frameworks are essential to prevent the proliferation of unregulated AWS and to ensure military AI advancements uphold fundamental human rights.

• Looking Ahead
  The global debate on regulating military AI continues. In April 2024, the Vienna Conference on Autonomous Weapons Systems gathered over 1,000 participants from 144 states to address the challenges and implications of lethal AWS. Building on his 2023 New Agenda for Peace, UN Secretary-General António Guterres has called for states to adopt a legally binding instrument by 2026. This treaty aims to ban lethal AWS that operate without human oversight and establish regulations for all other types of autonomous weapons systems.

Have something to add? Share your thoughts and help shape the discussion.

Leave a comment

Here are the results from the previous edition for this segment:

The poll results show that Upskilling Workforce for AI and Building Specialized AI Teams lead at 33% each. These responses highlight a growing recognition of the need for employees to adapt to AI-driven workflows and technologies, and the importance of establishing specialized teams to lead and scale AI integration efforts responsibly across departments.

Following closely, AI Transformation is Challenging at 22%, highlighting the broader difficulties organizations face, such as aligning technological initiatives with business objectives and overcoming resistance to change. A smaller but notable group of respondents are incorporating AI ethics education, acknowledging the importance of embedding ethical considerations into AI adoption.

As we highlighted in our AI and Labor Rights 2024 Review, these trends align with the reality that employers must take active steps to help their workforce adapt to AI advancements. Supporting job security and fostering equitable growth will be essential to navigating these transitions effectively.

How does your organization handle accountability for AI-driven decisions?
Do you assign ownership to specific teams or roles, use external audits and evaluations, rely on automated transparency features in AI systems, or currently have no formal accountability measures in place?

Share your thoughts with the MAIEI community:

Leave a comment

Mapping the Ethics of Generative AI: A Comprehensive Scoping Review

This comprehensive review synthesizes recent discussions on the ethical implications of generative AI, especially large language models and text-to-image models, using a scoping review methodology to analyze the existing literature. It outlines a detailed taxonomy of ethical issues in the domain of generative AI, identifying 378 distinct codes across various categories and highlighting the discipline’s complexity and the potential harms from misaligned AI systems. The research not only fills a gap by providing a structured overview of ethical considerations of generative AI but also calls for a balanced assessment of risks and benefits, and serves as a resource for stakeholders such as scholars, practitioners, and policymakers, guiding future research and technology governance.

To dive deeper, read the full summary here.

Careless Whisper: Speech-to-text Hallucination Harms

OpenAI’s speech-to-text service, Whisper, hallucinates entire sentences in addition to producing otherwise accurate speech transcriptions. These hallucinations induce concrete harms, including (a) perpetuating violence, (b) claiming inaccurate associations, and (c) projecting false authority. We find these harms to occur more frequently for speech with longer “non-vocal” durations (e.g., speech with more pauses or disfluencies), as evidenced by disproportionate hallucinations generated in our data among speakers with a language disorder, aphasia.

To dive deeper, read the full summary here.

AI Framework for Healthy Built Environments

How do we safeguard people’s health in built environments where AI is adopted? Research led by the International WELL Building Institute (IWBI) and Kairoi sets out a framework for built environment sectors to deploy and adopt AI in ways that are beneficial for people’s health and well-being.

To dive deeper, read the full summary here.

Microsoft expects to spend $80 billion on AI-enabled data centers in fiscal 2025 – CNBC

1. What happened: Microsoft plans to spend $80 billion in fiscal year 2025 on constructing data centers (over half of which will be in the US) to help handle the increased demands of AI technologies.

2. Why it matters: Increased pressure on the US to maintain its leadership in the AI space over countries such as China means that US companies have to invest heavily in critical infrastructure, such as data centers. Consequently, Microsoft and its competitors will increase the environmental toll these data centers require, such as water, electricity, and heat.

3. Between the lines: This significant financial commitment sends a strong signal to other competitors and nations that Microsoft and the US are doubling down further on AI. Additionally, this commitment pays homage to the current belief that for AI models to perform better, they need more data, which requires more data centers, despite disagreements over whether Big Tech is suffering a slowdown in the progress made by its newest iterations of its various models.

To dive deeper, read the full article here.

Meta scrambles to delete its own AI accounts after backlash intensifies – CNN

1. What happened: Meta faced backlash and removed several AI-generated user accounts, including “Grandpa Brian” and “Liv,” after public outrage over their misleading and stereotypical representations. These accounts were designed to drive engagement on Instagram and Messenger, but users discovered problematic inaccuracies, ethical concerns, and potential manipulation in their design.

2. Why it matters: The incident highlights critical AI ethics concerns, including the transparency of AI-driven content, the reinforcement of harmful stereotypes, and the manipulation of user trust. By deploying AI-generated personas that blurred fiction and reality, Meta risks undermining trust in its platform and raising questions about accountability for AI-generated content.

3. Between the lines: Meta’s AI personas were not just technical experiments—they were crafted to enhance engagement and monetization through emotional manipulation. The company’s prioritization of profit over ethical considerations and user trust reveals deeper systemic issues in how Big Tech integrates AI into its platforms. This incident raises broader concerns about corporate responsibility in AI deployment, transparency, and the potential societal impact of AI-generated narratives.

To dive deeper, read the full article here.

Don’t panic: AI can strengthen democracy too – Bulletin of the Atomic Scientists

1. What happened: This article explores how AI can strengthen democracy by improving governance, enhancing civic engagement, and ensuring fairer decision-making. While AI is often viewed as a threat to democratic values, this article highlights its potential to bolster transparency, combat disinformation, and create more equitable policy outcomes when used responsibly. AI-powered algorithms could more effectively elevate neutral, factual information, potentially countering common misconceptions on both sides of the aisle by analyzing vast amounts of data and presenting unbiased summaries.

2. Why it matters: AI’s role in shaping democratic processes has sparked intense debate. Unchecked AI risks exacerbating polarization, surveillance, and misuse by authoritarian regimes. However, when deployed with proper safeguards, AI can be a force for good—enhancing citizen participation, improving public services, and fostering accountability. This perspective is vital as nations grapple with balancing AI’s benefits against its risks to democratic systems.

3. Between the lines: The article urges policymakers to take proactive steps, including implementing ethical guidelines, ensuring human oversight, and addressing AI’s unintended consequences. Collaboration between governments, civil society and industry is essential to harness AI’s positive potential while mitigating threats to privacy, equity, and human rights.

To dive deeper, read the full article here.

What do we mean by the term “hallucination”?

👇 Learn more about why it matters in AI Ethics via our Living Dictionary.

Explore the Living Dictionary!

On with Kara Swisher Podcast: AI Ethics and Safety – A Contradiction in Terms?

Rumman Chowdhury, Mark Dredze, and Gillian Hadfield join Kara Swisher to discuss the complexities of AI ethics and safety. The panel answers questions like: is it possible to create unbiased AI? What are the worst fears and greatest hopes for AI development under Trump 2.0? What sort of legal framework will be necessary to regulate autonomous AI agents? And is the hype around AI leading to stagnation in other fields of innovation? Recommended listening for those navigating the ethical and societal implications of AI in 2025.

To listen to the full episode, click here.

South Korea’s National Assembly introduces its AI Basic Act

South Korea’s National Assembly has introduced a comprehensive AI Basic Act, consolidating fragmented regulations into a unified legal framework. This landmark legislation marks the first step toward creating a structured, forward-looking approach to regulating AI in the country.

To learn more about how the AI Basic Act compares to the EU AI Act, click here.

Does Military AI Have Gender? Understanding Bias and Promoting Ethical Approaches in Military Applications of AI

This research summary examines how theories of human-machine interactions can be strengthened by attending to the multiple ways humans are embodied. It draws on interdisciplinary research studying race and gender bias in commercial AI. The report argues ethical approaches to military applications of AI must be expanded by making transparent how gender, race, age and ability will be both explicitly and implicitly encoded in machine learning systems in development for national security. It also warns of the risks associated with increasing overlap between commercial and military applications of AI.

To dive deeper, read the full article here.

We’d love to hear from you, our readers, about any recent research papers, articles, or newsworthy developments that have captured your attention. Please share your suggestions to help shape future discussions!

Leave a comment