PDF Summary:Atlas of AI, by Kate Crawford

Atlas of AI by Kate Crawford offers an eye-opening exploration of artificial intelligence's profound impact beyond its technical boundaries. The book delves into AI's extensive environmental footprint involving resource extraction, manufacturing, and waste disposal, as well as its societal effects, including the exploitation of labor forces, perpetuation of biases, and potential erosion of privacy and democratic rights.

In this comprehensive analysis, Crawford examines the complex systems and intricate power dynamics intertwined with AI development and deployment. She sheds light on the practices of data collection, classification, and how algorithms reinforce existing inequalities, ultimately revealing AI's far-reaching consequences across diverse sectors of human life and the planet itself.

(continued)...

Practical Tips

• You can support the well-being of content moderators by using social media more responsibly. Before posting or sharing content, consider its potential impact on the individuals who may have to review it for appropriateness. This mindfulness can contribute to a healthier online environment and reduce the volume of harmful content that moderators have to sift through, potentially alleviating some of their stress.
• Start a mindfulness routine to prepare for the mental demands of working with AI systems. Mindfulness can increase your mental resilience and focus, which are crucial when dealing with repetitive tasks. You could use a mindfulness app that guides you through daily exercises aimed at improving concentration and reducing stress, thereby equipping you with the mental tools to better handle the taxing aspects of AI work.
• Start a conversation with your local representatives about the importance of safe working conditions in warehouses. By expressing your concerns and advocating for stricter regulations on workplace safety, you can contribute to creating a safer environment for warehouse employees.
• Develop a peer support network within your community or organization for individuals in high-stress jobs, including content moderation. This network could offer a safe space for sharing experiences and coping strategies. Start by reaching out to colleagues or community members to gauge interest, and then organize regular, informal meet-ups where people can connect and support each other.
• You can streamline your daily tasks by using a time-tracking app to identify inefficiencies. By monitoring how you spend your time at work, you can pinpoint activities that take up more time than they should. For example, if you find that email management is eating up hours of your day, you might decide to batch-process emails at set times instead of constantly checking your inbox.
• Introduce a personal policy of controlled exposure to distressing content by setting clear boundaries on when and how you engage with such material. Determine specific times of day when you're more resilient to stress and limit exposure to graphic content to these periods. For example, avoid confronting disturbing material right before bed to prevent it from affecting your sleep. Instead, choose a time when you can follow it with a positive or neutral activity, helping to buffer the negative emotional impact.

Algorithmic management has escalated the monitoring and control aspects, reducing the human interaction that was once a hallmark of workplace oversight.

Crawford argues that AI systems serve not only as replacements for human workers but also as instruments for monitoring and controlling employees in a manner that deeply invades their privacy and diminishes their self-esteem. The author illustrates how tools grounded in algorithmic technology are utilized across multiple sectors to organize work schedules, evaluate staff performance, and enforce disciplinary actions, all reflecting a uniform logic.

Crawford highlights the impact of AI on human autonomy and agency in the workplace, arguing that these technologies effectively turn workers into "appendages" of machines constantly responding to algorithmic demands. She argues that the focus on efficiency often obscures how these systems can exacerbate existing inequalities and reinforce historical patterns of labor exploitation.

The deployment of AI-driven surveillance of workers, which includes real-time performance monitoring and algorithmic scheduling, among other methods.

Crawford delves into the ways in which AI technologies intensify the monitoring of workplaces through the use of tools that meticulously observe and evaluate workers with unprecedented accuracy. Various technological instruments are employed, including devices that monitor physical movements and software that scrutinizes typing patterns and email interactions. In these environments, employees become mere data points, perpetually evaluated and compared based on metrics, with scant room for personal differences or autonomy.

Crawford highlights the significant imbalances in power, observing that workers often have minimal choices when it comes to their interaction with these technologies. The author connects the evolution of surveillance in industrial environments to its historical roots, emphasizing that the primary goals remain consistent over the ages: to improve supervision, increase efficiency, and extract economic benefits from the labor force.

Practical Tips

• Develop a habit of regularly reviewing and adjusting your privacy settings on work devices and applications. As new updates and features roll out, they may change the level of access your employer has to your activity. By staying proactive, you can maintain a level of personal privacy that aligns with your comfort and the expectations of your workplace.
• Experiment with free typing analysis tools available online to gain insights into your typing habits. These tools can provide you with data on your typing speed, accuracy, and common errors. Use this information to set personal benchmarks and improve your typing skills, which can enhance your efficiency in computer-based tasks.
• Develop a peer recognition system with your colleagues where you acknowledge each other's strengths and efforts that aren't captured by standard metrics. This could be as simple as sending a weekly email shout-out to a coworker or leaving a note of appreciation for someone's help or creative input. By doing this, you foster an environment that values personal differences and autonomy.
• Create a personal action plan for technology empowerment. Identify one piece of technology that you frequently use but feel disempowered by. Research alternatives or ways to customize it to better suit your needs. Implement one change and track its impact on your sense of control and job satisfaction over a month.

The perpetuation of existing disparities is fueled by the incorporation of biased and discriminatory methods into systems powered by artificial intelligence.

Crawford explores how AI systems could potentially exacerbate biases and discriminatory behaviors already present in the workplace instead of operating as neutral tools. The author emphasizes the notorious resume-sorting algorithm of a prominent e-commerce company to illustrate that depending on past data when creating training models can inadvertently perpetuate existing inequalities, including the underrepresentation of women in technical roles that require specialized skills.

Crawford suggests that certain demographic groups, especially those identified by their economic status, might continue to face biases because these systems often favor characteristics common among historically dominant groups. She emphasizes the need for greater awareness about how these algorithmic systems can perpetuate historical patterns of discrimination, even when designers have good intentions.

Practical Tips

• You can evaluate the fairness of AI by using online tools designed to detect bias in algorithms. Websites like AI Fairness 360 by IBM offer resources where you can input data and receive an analysis of potential biases. This helps you understand how biases manifest in AI and encourages you to think critically about the technology you use daily.
• Start a virtual book club focused on readings that highlight the achievements of women in STEM and discuss the challenges they face. This can be a space for both men and women to learn and reflect on the underrepresentation issue. You might select a book like "Hidden Figures" by Margot Lee Shetterly and organize monthly Zoom discussions to delve into the themes and real-world implications of the stories.
• Create a diverse media diet to challenge algorithmic biases in your content consumption. Actively seek out and engage with content from a variety of sources, especially those that represent different perspectives and communities than your own. This can include following creators from diverse backgrounds on social media, subscribing to newsletters from different parts of the world, or using a news aggregator that pulls from multiple outlets.

The ramifications of collecting data, its organization, and the built-in biases present in artificial intelligence algorithms.

The book delves into the intrinsic influence that training data has on AI systems, highlighting the inherently political aspects of gathering data, its classification, and the difficulties involved in addressing biases.

The progression of AI models has been driven by the extensive collection and assimilation of personal data, frequently without the individuals' permission.

Crawford argues that the norm within the technology sector involves amassing extensive personal data without consent, justifying this practice with the belief that more data is always better. She reveals the disconcerting methods of collecting vast amounts of data for training purposes, often derived from individuals' social media accounts and accessible public databases, with scant regard for the protection of personal information or the ethical handling of such data.

The author exposes the erroneous logic in considering data simply as an asset to be exploited, without recognizing that it encapsulates a collection of individual expressions, important life events, and confidential information. Crawford emphasizes the need to critically evaluate the drive to gather data and recognizes the potential harm that can result from constructing AI systems based on large-scale data assemblies that are gathered without consent.

The process of gathering and utilizing data from sources such as social media and government records is carried out lacking adequate transparency or oversight.

Crawford delves into the complexities that conceal the origins, classification, and utilization of data within the AI sector. She contends that by treating these techniques as trade secrets, companies restrict the sharing of details about data gathering, its uses, and the people employing it, which substantially impedes the capacity to assign accountability when harm or prejudice arises.

The author reveals collaborations among academic institutions, government entities, and private companies in their pursuit of data collection, frequently lacking proper oversight or openness, by providing examples like the DukeMTMC dataset case, where numerous images of students were employed to refine facial recognition technology unbeknownst to the individuals captured. Crawford emphasizes the need for improved supervision and transparency to ensure the responsible and ethical management of AI systems.

Other Perspectives

• Others might suggest that complete transparency could stifle innovation by making it harder for companies to invest in research and development if their findings are easily accessible to competitors.
• Transparency alone does not guarantee that the public or regulatory bodies will have the necessary expertise to understand and evaluate the data practices, which means that simply providing information is not enough without efforts to educate and empower stakeholders to use that information effectively.
• The process of establishing oversight and transparency is complex and may be in a state of evolution, with efforts underway to improve these aspects in response to public concern and regulatory changes.
• The dataset creators might have believed that the benefits of advancing facial recognition technology for security and safety purposes outweighed the privacy concerns.
• There is a possibility that increased oversight could lead to overregulation, which might inadvertently protect incumbent players and create barriers to entry for smaller companies or startups.

The invasion of individual privacy and liberty through the gathering of biometric data and the employment of systems engineered to identify individuals by their facial features.

Crawford examines the rapid increase in the collection of data driven by artificial intelligence, a trend that encroaches upon our essential rights to privacy and autonomy, especially focusing on the swiftly growing industry that specializes in facial recognition technology. The writer emphasizes the concerning absence of permission and the possibility for such systems to be utilized for monitoring and governing the public, referencing the case in which a repository of mug shots at NIST, which includes photographs of suspects and individuals who have passed away, is used for the development of facial recognition software.

The author further explores the dangers of deploying biased facial recognition tools in contexts like policing, border control, and public housing, arguing that these technologies can disproportionately harm marginalized communities. Crawford emphasizes the need for stronger legal and ethical frameworks to govern the collection and use of biometric data, particularly in public spaces, and argues that technologies capable of identifying individuals based on their facial features should be banned until adequate safeguards are in place.

Practical Tips

• Create a personal data inventory to keep track of where your information is stored. Make a list of all the online services and accounts you use, from social media to online shopping, and note what personal information you've provided to each. This will help you be more conscious of the data you share and allow you to take steps to remove or protect it if needed.
• You can protect your privacy by covering your webcam and phone camera when not in use to prevent unauthorized access and potential facial recognition misuse. Many devices can be hacked to access cameras without your knowledge, so physically covering your camera can be a simple yet effective barrier.
• Opt for alternative verification methods like two-factor authentication that don't rely on facial recognition to secure your online accounts.
• You can increase your awareness of facial recognition by opting into a service that notifies you when your image is detected in public databases or online platforms. This could involve signing up for a service that scans the internet for instances of your face and alerts you, allowing you to understand the extent of your public visibility and take action if necessary.
• Consider supporting organizations that advocate for ethical technology use by donating or spreading awareness. Your contribution helps promote the development and enforcement of ethical guidelines in technology. Look for non-profits or advocacy groups that focus on digital rights and privacy, and share their campaigns or materials on your social media to educate others.
• Start a habit of reading privacy policies for new apps and services before you agree to them. While it might seem tedious, getting into the routine of skimming for sections on biometric data will make you more informed about where and how your data might be used or shared.

The methods used to categorize data inherently possess political implications and frequently perpetuate harmful stereotypes along with existing power structures.

Crawford suggests that the categorization of data by AI systems mirrors and reinforces existing power structures, rather than operating as a neutral or objective mechanism. She argues that the creation of training datasets is not just a simple collection of raw data; rather, they reflect specific viewpoints, often incorporating inherent assumptions related to societal constructs like ethnicity, sex, and socioeconomic status.

Kate Crawford highlights the craniometry research conducted by Samuel Morton in the nineteenth century and contrasts these with contemporary artificial intelligence programs that classify individuals based on appearance, illustrating that attempts to quantify and classify purportedly objectively can perpetuate societal divisions and facilitate prejudiced practices. She underscores the necessity of acknowledging this pattern to prevent the exacerbation and continuation of historical injustices through mechanisms driven by AI.

The incorporation of racial, gender, and other categorical biases into the creation of training datasets and their corresponding classification mechanisms.

Crawford provides an in-depth analysis of the classification systems in widely used AI training datasets, emphasizing how such structures often perpetuate harmful stereotypes and exacerbate existing social inequalities. The author delves into datasets like UTKFace, revealing a reliance on basic and biased categorizations of gender, ethnicity, and personality traits, which serve to propagate a reductive and often discriminatory view of human variety.

Crawford suggests that the process of categorization within these systems reflects the perspectives of a limited and homogeneous group, primarily originating from prestigious educational institutions and powerful companies in the tech industry. She emphasizes the need to cultivate a critical awareness that expands diversity in the field of AI, which in turn helps to halt the continuation of harmful stereotypes about marginalized communities.

Other Perspectives

• The use of broad categories in training datasets can sometimes be a practical necessity for managing large amounts of data and may not always be intended to convey a reductive view of human variety.
• The use of open-source models and datasets can encourage contributions from a global community, potentially mitigating the influence of a limited and homogeneous group on the categorization process.
• The group behind these systems is not exclusively from prestigious educational institutions and powerful tech companies; there is a growing number of contributors from diverse backgrounds and smaller organizations involved in AI development.
• Critical awareness alone may not lead to the desired change in diversity if there is no clear action plan or policy to implement the changes. Concrete steps and accountability mechanisms are necessary to translate awareness into actual diversity in the workforce.
• The focus on increasing diversity might overlook the need for better education and training for all AI practitioners on the ethical implications of their work, regardless of their background.

Efforts to eliminate bias via technical solutions fail to address the deeper societal and political roots of prejudice embedded in artificial intelligence systems.

Crawford argues that although it's essential to implement technical strategies to reduce bias within artificial intelligence systems, these methods often fail to address the deeper societal and political roots of inequality. The writer highlights that while some efforts are well-meaning, like those intended to improve the variety of individuals recognized by facial recognition technology, they can inadvertently reinforce harmful stereotypes and erroneously consider traits like race and gender as unchangeable biological features.

The author stresses the importance of going beyond merely aiming for demographic diversity and instead tackling the fundamental characteristics that distinguish these groups, in order to move away from a technocentric viewpoint that overlooks the complex social and cultural context. To mitigate bias, it is crucial to scrutinize the underlying principles of categorization and address the broader systemic inequalities perpetuated by these technologies.

Context

• Technical solutions, such as algorithmic adjustments, often focus on surface-level corrections without addressing the root causes of bias, which are deeply embedded in social structures and historical contexts.
• Race and gender have been socially constructed over time, influenced by historical, cultural, and political factors, rather than being purely biological. This understanding challenges the notion of these categories as fixed.
• Addressing fundamental characteristics involves examining how power structures influence which traits are emphasized or marginalized in society. This requires a critical look at who benefits from maintaining certain categorizations and who is disadvantaged.
• Technocentrism is an approach that prioritizes technological solutions over social or cultural considerations. It often assumes that technology can solve complex human problems without addressing the underlying societal issues.
• The principles of categorization in AI often draw from historical practices of classification that have been used in various fields such as anthropology and biology, which have sometimes perpetuated stereotypes and biases.

The integration of artificial intelligence by the government into its systems of administration, regulation, and monitoring.

Crawford delves into the complex relationship between artificial intelligence and state power, scrutinizing how AI tools are employed not just for monitoring and controlling citizens, but also for reshaping the very nature of governance and supervision.

Intelligence and law enforcement agencies are progressively utilizing AI tools to enhance their surveillance capabilities and to implement proactive policing strategies.

Crawford explores how artificial intelligence expands and intensifies conventional methods of surveillance by governments across the globe. She delves into the secretive domain where intelligence agencies employ sophisticated algorithms to amass vast amounts of information, assess potential threats, and carry out targeted operations.

Crawford highlights the strategy employed by the NSA in creating initiatives like TREASUREMAP and FOXACID, which involves collecting extensive data sets to detect anomalies via analytical methods. She reveals how these approaches, once exclusive to intelligence agencies, have now filtered down to local police departments through partnerships with companies like Palantir, leading to increased surveillance of vulnerable communities, particularly immigrants, the undocumented, and communities of color.

The expansion of state monitoring and control over citizens through AI-driven facial recognition, license plate tracking, and other biometric identification

Crawford unveils the disconcerting evolution of traditional law enforcement practices, now augmented to enable broader and more proactive surveillance methods through the integration of artificial intelligence. She explores the widespread integration of technology by police forces that can recognize people through their facial characteristics, emphasizing its potential to hinder the quest for fairness and its tendency to have an adverse impact on communities that are often marginalized.

Crawford explores the implications of employing technologies like automated vehicle identification systems (AVIS) to monitor individuals and compile extensive databases that are available to a range of government and private organizations. She argues that these technologies are blurring the lines between traditional policing and intelligence gathering, leading to pervasive monitoring that can greatly suppress civil freedoms and dissent.

Practical Tips

• Enhance your digital privacy by using tools that obscure your facial features or license plates in photos before you post them online. Look for apps or software that can blur or pixelate specific areas of your images. By doing this, you reduce the risk of your personal images being used to train AI systems without your consent.
• Engage with community safety initiatives by proposing the use of AI technology to analyze public space usage patterns. This could involve collaborating with local authorities to implement AI tools that monitor foot traffic in high-crime areas, helping to predict and prevent potential incidents. For instance, if a particular park has a spike in visitors after dark, the AI could flag it for increased patrols during those hours.
• Opt for public transportation or rideshare services where personal vehicle identification isn't tied to you. By using these services, you're not directly subject to AVIS tracking in the same way as when driving your own vehicle. For instance, when you take a bus, train, or a rideshare, the vehicle's identification is linked to the service provider, not to you as an individual.
• You can enhance your privacy by using encryption tools for your digital communication. Since the lines between policing and intelligence gathering are blurred, protecting your personal information becomes crucial. Start by using end-to-end encrypted messaging apps like Signal or WhatsApp for your daily communications, and consider using a VPN service to secure your internet browsing activities.
• Engage in face-to-face conversations about sensitive topics instead of online communication. Meeting in person in a private setting reduces the risk of your conversations being monitored or recorded.
• Encourage open conversations with friends and family about the impact of state monitoring. Share articles, videos, or infographics that explain how surveillance can affect different communities, especially marginalized ones. Discussing these issues can raise awareness and potentially influence the opinions and voting choices of your social circle, contributing to a broader cultural shift.

The biased and discriminatory impacts of AI-based "risk assessment" tools in criminal justice and social welfare contexts

Crawford scrutinizes the concerning tendency to entrust AI systems with critical decision-making for people, even though these systems are prone to mistakes and partiality. She explores the utilization of artificial intelligence-driven instruments within the judicial system to predict the probability of reoffending and influence sentencing outcomes. She contends that such technologies may perpetuate preexisting prejudices, resulting in biased outcomes that particularly affect members of historically underrepresented communities.

Crawford explores the integration of artificial intelligence in the welfare system, revealing how automated decisions can restrict access to benefits and impose sanctions on the needy. She underscores the case where Michigan's MiDAS, an autonomous operating system, mistakenly accused many individuals of fraud and imposed severe penalties, showing a tendency in these systems to prioritize cost-cutting and fraud detection over providing support to those in need.

Practical Tips

• Start a journal to track and reflect on decisions made by AI in your daily life. Whether it's a recommendation from a streaming service or a navigation app suggesting a route, note the outcome and your feelings about the AI's choice. This practice will help you become more aware of how often AI influences your decisions and the level of trust you place in these systems.
• Volunteer with a local legal aid organization and ask to be involved in cases or projects that deal with AI risk assessments. Even without legal expertise, you could offer to help with research, documentation, or supporting community outreach programs that educate the public about how AI is being used in the legal system. This hands-on experience will provide you with a deeper understanding of the practical challenges and benefits of AI in this context.
• You can start a personal audit of your daily interactions to identify biases. Keep a journal for a month where you note down decisions you make that involve others, especially those from underrepresented communities. Review this journal weekly to spot patterns or biases in your behavior, and then actively work to adjust your actions to be more inclusive.
• Develop a habit of documenting all interactions with welfare systems. Keep a detailed log of application dates, phone calls, emails, and any automated notices received. This record can be invaluable if you need to contest a decision or prove compliance with welfare requirements. Use a simple spreadsheet or a dedicated notebook for this purpose, ensuring that you date each entry and include as much detail as possible.
• You can safeguard your rights by regularly reviewing your credit report and financial statements for inaccuracies. If you notice any discrepancies, especially related to fraud allegations or debts you don't recognize, take immediate action by contacting the credit bureaus and financial institutions to dispute the errors. This proactive approach helps you catch and address potential issues early, mirroring the importance of accuracy and vigilance in automated systems like MiDAS.
• Advocate for more human-centered AI by writing to local representatives or companies that use AI systems, suggesting they consider the impact on individuals in need. Highlight the importance of incorporating empathy and support into AI design, and propose that they seek input from diverse user groups to ensure the technology serves the community equitably.

The increasing integration of technology companies with defense and security systems is facilitating the use of artificial intelligence for military purposes.

Crawford argues that the line between private sector AI development and government use is becoming increasingly blurred, particularly when it comes to military and security applications. She investigates how corporations have embraced and widely implemented tools once solely in the domain of intelligence agencies, resulting in a concerning escalation of AI applications that resemble military activities.

Crawford highlights that companies like Palantir, initially founded to aid military intelligence, have broadened their scope to supply police, immigration officials, and corporate entities with systems for surveillance and analytical services for data. Critical questions arise about the integrity of legal processes, the accountability of democratic bodies, and the potential for these technologies to encroach on individual rights and privileges when boundaries become intertwined.

The Pentagon's strategy, known as the "Third Offset Strategy," focuses on incorporating artificial intelligence into its military functions and surveillance efforts.

Crawford scrutinizes the Pentagon's approach, which is to sustain military dominance by integrating artificial intelligence throughout various combat elements, including autonomous weapons, precision in drone strikes, and extensive surveillance activities. She explores the controversial partnership involving Google and the U.S. Department of Defense's Project Maven, which sought to develop systems capable of autonomously processing imagery from unmanned aerial vehicles, highlighting the turmoil within Google due to employee dissent against the company's involvement in defense-related initiatives.

Crawford explores the ethical implications of "signature strikes," where people are targeted based on behavioral resemblances rather than concrete evidence of their identity, and underscores the perils of allowing artificial intelligence to determine outcomes that could mean the difference between survival and fatality. Crawford contends that the fusion of the tech sector with military organizations has been driven by economic incentives, governmental pressure, and nationalistic sentiments, all of which collectively erode unbiased examination and transparent discussion regarding the ethical use of artificial intelligence.

Other Perspectives

• The strategy may divert attention and resources from other critical areas of defense, such as cyber security or human intelligence, which are also vital for national security.
• Precision in drone strikes, while technologically advanced, does not guarantee the absence of civilian casualties and can lead to unintended consequences, undermining the perceived precision.
• Google's work on Project Maven could be defended as part of the tech industry's role in supporting legal governmental activities, including defense and intelligence operations.
• The internal dissent at Google may not necessarily reflect the overall sentiment of the entire company but could represent the views of a vocal minority.
• "Signature strikes" may incorporate more than just behavioral resemblances; they often use a combination of intelligence sources, including human intelligence and signal intelligence, to corroborate the behaviors observed.
• AI can process vast amounts of data more quickly and accurately than humans, which can lead to more informed and precise decisions in life-threatening scenarios, potentially saving lives.
• Collaboration between the tech sector and military can be seen as a response to evolving security challenges, rather than purely economic incentives or governmental pressure.
• The integration of AI in military functions is often subject to internal and external oversight, suggesting that there is still room for ethical considerations and transparent discussions, even in the context of national security.

The lines between governmental authority and corporate influence are increasingly blurred with the growing clout of AI platforms.

Crawford highlights how technology companies are becoming more entwined with critical public responsibilities, such as administering social welfare programs and carrying out activities related to policing, blurring the lines between business interests and public governance. She argues that companies like Palantir, deeply engaged with ICE and similar agencies, essentially function as government appendages, yet they circumvent the strict scrutiny and openness typically demanded of governmental entities.

The growing trend of allowing private entities in control of artificial intelligence systems to influence governance raises significant issues about protecting democratic principles, ensuring transparency in processes, and the potential for abuse. Crawford suggests that without strict regulation and clear responsibility, there could be a decline in personal freedoms and a reduction in public confidence, which happens alongside the growing alliance between governmental entities and tech companies.

Other Perspectives

• The collaboration between technology companies and government can be seen as a form of outsourcing, which is a long-standing practice in public administration to leverage the specialized skills of the private sector.
• Companies like Palantir may argue that they are subject to a different set of regulations and scrutiny that are appropriate for private sector entities, which can include audits, compliance checks, and contractual obligations that ensure they meet certain standards when working with government agencies.
• Collaboration between government and private entities can lead to a cross-pollination of best practices, which can enhance the protection of democratic principles rather than erode them.
• Public confidence is not solely dependent on regulation and clarity of responsibility; it can also be influenced by the effectiveness of the services provided, public relations, and the perceived benefits of AI platforms.
• There are instances where tech companies have been penalized or scrutinized by government authorities, indicating that the relationship is more complex and sometimes adversarial, rather than a straightforward alliance.