Broken Promises and Unmet Commitments
The omission of a safety research report alongside the release of Gemini 2.5 Pro is widely viewed as a violation of prior assurances given by Google. Back in July 2023, Google actively participated in a White House summit, orchestrated by the Biden administration, where it joined forces with other prominent AI companies to endorse a series of pledges. A pivotal commitment involved the publication of comprehensive reports for all major public model releases that surpassed the current cutting-edge AI capabilities at that time. Given its notable advancements, Gemini 2.5 Pro would undoubtedly fall within the purview of these White House Commitments.
At that juncture, Google explicitly agreed that these reports should encompass several critical elements:
- Thorough safety evaluations conducted, encompassing comprehensive assessments of potentially dangerous capabilities.
- Identification of significant performance limitations that could substantially impact the appropriateness of various use cases.
- In-depth discussions of the model’s anticipated effects on prevailing societal risks, such as fairness and pervasive bias.
- Detailed results of rigorous adversarial testing, meticulously designed to rigorously evaluate the model’s overall fitness for real-world deployment.
Subsequent to the G7 summit convened in Hiroshima, Japan, during October 2023, Google and a consortium of other leading companies pledged to steadfastly adhere to a voluntary code of conduct specifically tailored for the responsible development of advanced AI systems. This G7 code underscored the paramount importance of publicly reporting advanced AI systems’ core capabilities, inherent limitations, and suitable as well as unsuitable applications. The overarching aim was to substantially enhance transparency and fortify accountability within the rapidly evolving AI field.
In May 2024, at a prestigious international summit on AI safety, which was held in Seoul, South Korea, Google unequivocally reiterated its unwavering commitments. The company pledged to publicly disclose all relevant model capabilities, inherent limitations, and appropriate and inappropriate use cases. Furthermore, Google committed to provide comprehensive transparency surrounding its risk assessments and the subsequent outcomes.
Google’s Response and Delayed Transparency
In response to mounting inquiries regarding the conspicuous absence of the crucial safety report, a spokesperson representing Google DeepMind, the specialized division responsible for spearheading the development of the groundbreaking Gemini models, asserted that the latest iteration of Gemini had undergone extensive pre-release testing. This rigorous testing encompassed thorough internal development evaluations and comprehensive assurance evaluations that were meticulously conducted prior to the model’s highly anticipated public release. The spokesperson also indicated that a detailed report containing additional safety information and comprehensive model cards was “forthcoming.” However, despite the initial statement being officially issued on April 2nd, no model card has been officially published to date, raising further concerns about transparency.
A Wider Trend of Neglecting Safety Reporting
Google is not an isolated case in facing mounting criticism regarding its unwavering commitment to AI safety. Earlier this year, OpenAI also encountered significant scrutiny for its perceived failure to release a timely model card for its highly anticipated Deep Research model. Instead, they opted to publish a system card several weeks after the project’s initial release, raising eyebrows among industry observers. Similarly, Meta’s recent safety report for Llama 4 has been widely criticized for being excessively brief and conspicuously lacking in essential detail.
These recurring instances collectively highlight a deeply concerning trend within the AI industry, where some of the major labs are seemingly failing to prioritize safety reporting in strict accordance with their respective model releases. This emerging pattern is particularly troubling given the voluntary commitments that these companies willingly made to the U.S. government and the broader global community to diligently produce such comprehensive reports. These commitments were initially made to the Biden administration back in 2023 and were subsequently reinforced through formal pledges to steadfastly comply with the AI code of conduct that was formally adopted by the G7 nations at their AI summit in Hiroshima.
Kevin Bankston, a highly respected advisor on AI governance at the esteemed Center for Democracy and Technology, expressed his profound disappointment, stating unequivocally that these failures undermine the credibility of the companies involved and raise serious questions about their genuine commitment to responsible AI development.
Unanswered Questions and External Evaluation
The official statement issued by the Google spokesperson also conspicuously failed to address several specific questions regarding whether Gemini 2.5 Pro had been formally submitted for rigorous external evaluation by either the U.K. AI Security Institute or the U.S. AI Safety Institute. Previously, Google had voluntarily provided earlier generations of its groundbreaking Gemini models to the U.K. AI Safety Institute for comprehensive evaluation.
At the Seoul Safety Summit, Google formally signed up to the “Frontier AI Safety Commitments,” which included a formal pledge to provide public transparency on the thorough implementation of safety evaluations. The only explicitly stated exceptions were cases where doing so would demonstrably increase the overall risk or divulge highly sensitive commercial information to a degree that would be disproportionate to the societal benefit. The pledge also clearly stated that more detailed information that could not be publicly shared should still be shared directly with the governments of the countries in which the companies are based, which would be the U.S. in Google’s specific case.
The participating companies also unequivocally committed to explain precisely how external actors, such as governments, civil society organizations, prominent academics, and the broader public, are actively involved in the meticulous process of assessing the inherent risks associated with their sophisticated AI models. Google’s apparent failure to answer direct questions regarding whether it has indeed submitted Gemini 2.5 Pro to either U.S. or U.K. government evaluators potentially violates this fundamental commitment as well.
Prioritizing Deployment Over Transparency
The conspicuous absence of a comprehensive safety report has invariably raised significant concerns that Google may be currently prioritizing rapid deployment over the crucial elements of transparency and thorough safety assessments. Sandra Wachter, a highly regarded professor and senior researcher at the prestigious Oxford Internet Institute, emphatically emphasized the paramount importance of transparency in responsible research and innovation. She astutely drew a compelling analogy to other well-established industries, stating that “If this was a car or a plane, we wouldn’t say: let’s just bring this to market as quickly as possible and we will look into the safety aspects later.” Wachter further expressed her deep concern that there appears to be a prevailing attitude within the rapidly evolving generative AI field of “putting this out there and worrying, investigating, and fixing the issues with it later.”
Political Shifts and Competitive Pressures
Recent political shifts, coupled with the escalating rivalry among Big Tech companies, may be contributing to a discernible shift away from previously held safety commitments as these companies engage in a relentless race to deploy cutting-edge AI models. Wachter keenly noted that “The pressure point for these companies of being faster, being quicker, being the first, being the best, being dominant, is more prevalent than it used to be,” adding that safety standards were demonstrably declining across the entire industry.
These slipping standards could potentially be driven by a growing concern among tech countries and some governments that existing AI safety procedures are actively impeding overall innovation. In the U.S., the Trump administration has explicitly indicated its intention to adopt a less stringent approach to AI regulation compared to the more cautious stance taken by the Biden administration. The new administration has already rescinded a Biden-era executive order on AI and has been actively cultivating closer relationships with prominent tech leaders. At the recent AI summit in Paris, U.S. Vice President JD Vance stated unequivocally that “pro-growth AI policies” should be prioritized over safety, and that AI was “an opportunity that the Trump administration will not squander.”
At the same summit, both the UK and the U.S. notably declined to sign an international agreement on artificial intelligence that explicitly promoted an “open,” “inclusive,” and “ethical” approach to the technology’s development, raising further questions about their commitment to global AI safety standards.
The Need for Clear Transparency Requirements
Bankston emphasized that “If we can’t count on these companies to fulfill even their most basic safety and transparency commitments when releasing new models—commitments that they themselves voluntarily made—then they are clearly releasing models too quickly in their competitive rush to dominate the field.” He added that as AI developers continue to falter in upholding these critical commitments, it will be incumbent on lawmakers to proactively develop and vigorously enforce clear transparency requirements that the companies cannot readily shirk.
The Broader Implications for AI Governance
The controversy surrounding Google’s Gemini 2.5 Pro and the conspicuously missing safety report underscores the critical need for robust AI governance frameworks. These frameworks should comprehensively address key issues such as:
- Transparency: Ensuring that AI developers are consistently transparent about the capabilities, inherent limitations, and potential risks associated with their cutting-edge models.
- Accountability: Establishing clear and well-defined lines of accountability for the responsible development and ethical deployment of AI systems.
- Safety: Implementing rigorous safety testing and comprehensive evaluation procedures to effectively minimize the overall risk of potential harm.
- Ethical considerations: Integrating fundamental ethical principles into the very design and development processes of AI systems.
- Public engagement: Engaging proactively with the broader public to foster a more comprehensive understanding of AI and its wide-ranging implications.
- International cooperation: Collaborating internationally to develop common standards and universally accepted best practices for effective AI governance.
The conspicuous lack of transparency surrounding Gemini 2.5 Pro effectively highlights the potential consequences of neglecting these crucial aspects of responsible AI governance. Without adequate transparency and strict accountability, it becomes exceedingly difficult to accurately assess the true impact of AI systems and to ensure that they are developed and deployed in a consistently responsible and ethical manner, thereby safeguarding public trust and promoting societal well-being.
Moving Forward: A Call for Greater Responsibility
The AI industry currently stands at a critical juncture. As AI technologies become increasingly powerful and pervasive, it is absolutely essential that developers prioritize safety, transparency, and ethical considerations above all else. The ongoing controversy surrounding Gemini 2.5 Pro serves as a stark reminder that voluntary commitments, while laudable, are simply not always sufficient. Governments and regulatory bodies must proactively play a much more active role in establishing clear and enforceable standards and ensuring consistent compliance across the board.
Moreover, it is absolutely crucial for AI developers to actively engage with the public and proactively foster a broader understanding of AI and its vast implications. This includes being consistently transparent about the inherent limitations and potential risks associated with AI systems, as well as diligently communicating the specific steps being actively taken to effectively mitigate those risks. By working collaboratively, the AI industry, governments, and the public can collectively ensure that AI technologies are developed and responsibly deployed in a manner that genuinely benefits society as a whole, fostering innovation while safeguarding against potential harm.