A Vanishing Act: What We Can Learn From Google’s Decision to “Disappear” Its Principles Page
By Deb Donig, All Tech Is Human’s Siegel Research Fellow
“This shift has implications for both how companies direct their energies and what technologies ultimately get built.”
The period between 2016-2022 was marked by a particular refrain of corporate responsibility in the tech sector. Responding to the 2016-2017 “techlash,” or backlash against the harms caused by the development of tech products that willfully ignored, or that was unconcerned with, the negative consequences of tech products, companies made prominent public statements and commitments to pursuing public good, and to demonstrating the ways in which they were actively practicing and directing resources toward mitigating harms. It was in that moment that Google, a company launched under the banner “don’t be evil,” published its principles page.
If you have tried to access the link to that page that I have supplied in this article, you’ll notice that the link leads to a PDF, and not an actual webpage. That is because last week Google removed that page, erasing the principles as part of its public stance and presumably reneging on its promise to hold itself to the ethical standards and commitments it had proposed, among them the commitments to 1) be socially beneficial; 2) avoid creating or reinforcing unfair bias; 3) be built and tested for safety; 4) be accountable to people; 5) incorporate privacy design principles; 6) uphold high standards of scientific excellence; 7) be made available for uses that accord with these principles. In recanting these principles, Google also withdrew its commitment to not pursuing certain applications of AI, including 1) creating technologies that cause or are likely to cause harm; 2) creating technologies principally designed as weapons or facilitating injury to people; 3) that use information for the purpose of surveillance against accepted norms; 4) whose purpose contravenes international law and human rights.
It is unclear whether the decision to remove the webpage was motivated by unharnessing Google from its obligation to the first set of principles, which positively frame what Google will do, or whether it is rather motivated by a decision by Google to untether itself from its previously stated obligation to the second set of principles, which negatively frame what Google will not do. Both are concerning. But the two sets of principles are pursuant of two different sets of conditions.
We might restate these two sets of principles in terms of rights, and entitlements. Entitlements are, in general, things that we are owed – typically by governments – that we are obligated to receive as benefits of living in a society. Rights are, in contrast, things that we are generally protected from – again, typically from governments. We have an entitlement to education in the United States, because the government has ensured provisions for persons under the age of 18 to access education, and mandated that persons under 18 go to school. But the United States cannot, say, enslave people and force them to teach students; the right to education cannot supercede the right of people to decide whether or not to take up the job of teaching people’s children. Conversely, rights are protections of citizens from things that a government or power could do to them. We have a right to not be detained or have our houses searched without due cause. Authorities cannot arrest or detain us without a legitimate legal reason, without following proper legal procedures, and without providing the detained person access to courts to challenge their detention. These are protections that acknowledge the power of those authorities to intrude into the lives of citizens, and that restrain those authorities from doing so – this set of rights, called “habeus corpus” literally means “produce the body.” The government cannot trespass into the lives of its citizens in this way without first producing the body, and establishing just cause, thus limiting the scope of that power over citizens substantially.
This difference between rights and entitlements allows us to see the difference between the first set of principles in Google’s AI policy, and the second. The first act as a set of investments by Google toward developing certain practices and performances.
These principles suggest that Google would have, pursuant to those principles, had to spend more resources in cultivating technologies that are socially beneficial, that avoid creating or reinforcing unfair bias, that are built and tested for safety, and that are accountable to people, etc. The second set of principles, however, represent categories of things Google had previously asserted that it will not invest in developing – weapons, surveillance, etc. Establishing the categorical difference between rights and entitlements allows us to see the policy shift for what it really is – a restructuring of financial resources around driving incentives and profit directives. Google's recent removal of its AI weapons prohibition marks a concerning shift in corporate AI ethics. The original 2018 AI principles explicitly stated Google wouldn't develop AI for weapons or surveillance. Removing these commitments suggests increasing pressure to monetize AI capabilities regardless of ethical implications.
This change reflects broader tensions between commercial interests and responsible AI development. As AI capabilities advance, military and surveillance applications become more lucrative. The Department of Defense's $9 billion cloud computing contract with Google and other tech companies demonstrates the financial incentives driving these decisions; the Trump administration’s announcement that it is committing $500 billion in AI investments makes the combination of technologies and tech companies that can combine these capabilities makes this direction a seductive one for companies.
Tech companies originally adopted ethical AI principles to build public trust and address employee concerns. Google's 2018 principles emerged after employee protests over Project Maven, which used AI for military drone surveillance. The reversal indicates corporate priorities may increasingly favor government contracts and shareholder value over ethical restraints. Moreover, it reduces the incentives for workers, and the perceived power of those workers, to enact and advocate for their values in the workplace. This sends a powerful signal to employees: value-driven work or beliefs are unwelcome in the workplace. The change also reveals limitations of voluntary corporate AI principles. Without binding regulation or external enforcement, companies can modify or abandon ethical commitments as market incentives shift. This suggests effective AI governance may require legislative frameworks rather than relying on corporate self-regulation.
This shift has implications for both how companies direct their energies and what technologies ultimately get built. The directives provided by these incentives could accelerate the militarization of AI technology. Without clear corporate boundaries, advanced language models, computer vision, and robotics may be repurposed for autonomous weapons and mass surveillance. Other tech companies may follow Google's lead to remain competitive for defense contracts.The timing is particularly significant given rapid advances in foundation models and robotics. As these technologies become more capable, their potential military applications expand, and theteh combination of growth across these two industries in particular, coupled with the financial incentives for developing them in the form of government contracts, provides new incentives to their development.
The question remains as to whether Google's decision may also impact talent recruitment and retention. It is true that a driving force behind the decision to pursue a career in the tech industry is economic in nature, as these jobs offer some of the most lucrative careers, particularly in areas where the cost of living is high. Still, it is also true that a not insignificant number of high-level advanced AI researchers and engineers choose employers based on ethical commitments, values, and a wish to avoid working at companies that invite public controversy and backlash. Google, Meta, and other companies have previously experienced employees leaving their companies because of ethical or ideological disagreements with the company, and the stances taken by these companies will have an impact on the sense of value, autonomy, and job enjoyment workers experience. With the constriction in the market, this may have less of an impact than in previous historical moments where workers have left for ideological differences, as employees have fewer alternative options and experience the reality of job precarity, but employees, especially in-demand talent, may start to migrate toward other opportunities.
Overall, however, Google's decision is emblematic of the reality that AI development has determinatively shifted away from an era of aspirational principles, toward market-driven development. Following Open AI’s overt abandonment of its previous commitments to public interest and civically driven endeavors, animated by research curiosity and public commitments to public good in 2024, major companies are moving forward with agendas that no longer feel the need to even pay lip service to developing digital technologies in the service of principles.
Google's reversal demonstrates that ethical AI development requires more than voluntary corporate principles - it needs sustained commitment from multiple stakeholders and potentially legal frameworks to ensure responsible innovation. We are seeing – in real time – that frameworks for responsible development that prioritize self-governance or the development of ethical principles are insufficient; if they can be overturned overnight by simply deleting an active webpage, they never really were constraints in the first place. This moment should alert us to the necessity of regulation and legally codified rules of engagement, not principles self-determined by companies that operate on an opt-in basis.
This moment demands broader public discourse about acceptable uses of AI technology and mechanisms to enforce boundaries consistent with the principles of civil rights and liberal democracy. Of course, this is contingent upon a functioning legislative branch whose aims and operations are consistent with those same principles. Whether we have this at present remains to be seen. In the meantime, this is an opportunity for civil society to start building on existing movements to add social pressure and consensus to what we want our technologies to build toward. While commercial pressures may push toward unrestricted development, our society should start envisioning what restrictions we want to activate – not as goodwill corporate strategy, but as future laws.
“This moment demands broader public discourse about acceptable uses of AI technology and mechanisms to enforce boundaries consistent with the principles of civil rights and liberal democracy.”
==