Responsible AI Impact Report: Urgent risks, emerging safeguards, and public interest solutions impacting society
All Tech Is Human has just released in inaugural Responsible AI Impact Report, which is a crucial roadmap of the most urgent risks, emerging safeguards, and public-interest solutions shaping how AI will impact society in the year ahead.
All Tech Is Human’s Responsible AI Impact Report provides an assessment of the global Responsible AI (RAI) landscape, highlighting some of the most influential civil society contributions that shape AI governance, assurance, safety, and public-interest infrastructure. It documents a RAI field in transition; from principles to practice, from voluntary guidelines to enforceable standards, and from private AI dominance toward emerging visions of Public AI designed for collective benefit.
The Responsible AI Impact Report outlines how the Responsible AI ecosystem has matured into a more coordinated, evidence-driven, and public-interest–oriented field. Civil society organizations play a central role in shaping regulatory debates, expanding global governance frameworks, and surfacing real-world harms across safety, security, privacy, fairness, labor, climate, and democratic integrity.
The report emphasizes that frontier and agentic systems are amplifying existing risks while simultaneously exposing the inadequacy of current safeguards. The report examines areas such as biosecurity uplift and data poisoning to synthetic media manipulation, fraud, biased biometrics, and the potential psychological harms of AI companions. We highlight the community’s growing emphasis on rigorous AI assurance, including interoperable standards, lifecycle documentation, community-aligned benchmarks, independent audits, and participatory red teaming that reflects diverse global contexts.
Our report looks to 2026 with clear priorities: strengthening rights-anchored regulation amid deregulatory pressures; scaling assurance practices that generate usable, audit-ready evidence; closing capacity gaps for nonprofits and public agencies; safeguarding information integrity; guiding synthetic companion governance; and expanding shared Public AI infrastructure.
At stake is the defining question for the years ahead: Who determines the purpose of AI and the kinds of lives it will shape?"
Who should read the Responsible AI Impact Report?
AI governance practitioners in industry
Civil society organizations
Researchers and academics
Government regulators and policymakers
Public-interest technologists
Educators, students, and RAI newcomers
Philanthropic funders supporting Responsible Tech
Feedback | Reach out to All Tech Is Human
The lead author for the Responsible AI Impact Report is Rebekah Tweed. Strategic and editorial support for the report was provided by David Ryan Polgar, Sandra Khalil, and Sherine Kazim. The report was produced by the non-profit organization All Tech Is Human. All Tech Is Human is an organization developing a better approach for tackling thorny tech & society issues; one that leverages collective intelligence, involvement, and action to create needed systemic change.
The Responsible AI Impact Report received contributions and feedback from Alisar Mustafa, Jen Weedon, Justin Hendrix, Leah Ferentinos, Merve Hickok, Michelle Shevin, Dr. Nathan C. Walker, and Theodora Skeadas. The report features submitted quotes from Alexis Crews, Avijit Ghosh, Borhane Blili-Hamelin, Claire Leibowicz, Dhanya Lakshmi, Dr. Joy Buolamwini, Kevin Barenblat, Madhulika Srikumar, Merve Hickok, Miranda Bogen, Tamara Kneese, and Vilas Dhar. (Read their quotes below)
Some of the organizations cited in our report include the Ada Lovelace Institute, AI Now Institute, The Alan Turing Institute, Algorithmic Justice League, All Tech Is Human, Aspen Digital, Atlantic Council (DFRLab), Better Images of AI, Center for AI and Digital Policy, Center for Democracy & Technology, Centre for the Governance of AI, Coalition for Content Provenance & Authenticity, Collective Intelligence Project, Common Sense Media, Consumer Reports, Data & Society, Distributed AI Research Institute, Equal AI, Estampa, EvalEval Coalition, Fast Forward, Federation of American Scientists, Forecasting Research Institute, The Future Society, Humane Intelligence, Information Professionals Association, Japan AI Safety Workshop, Knight First Amendment Institute at Columbia, Metrology Working Group, MIT AI Risk Initiative, MITRE, MLCommons, Montreal AI Ethics Institute, Mozilla Foundation, The OpenFold Consortium, Patrick J. McGovern Foundation, Partnership on AI, Project Liberty, Public Interest AI Project, Reboot, Responsible AI Collaborative, Singapore AI Safety Institute, The Stanford Institute for Human-Centered AI, The Tech We Want, UNESCO, Weval, and WITNESS.
“An AI-enabled future grounded in human dignity depends on institutions that can govern powerful technologies with a commitment to the public good. When societies build this kind of civic architecture, people gain the ability to direct technological development rather than be shaped by it. This report highlights how that architecture is emerging through civil society’s work: accountable standards, rigorous evaluations, provenance systems that reinforce information integrity, and safeguards that respond to prominent harms. Together, these efforts outline a path toward a digital world that strengthens democratic agency and reflects the values we choose to uphold.”-Vilas Dhar, President, The Patrick J. McGovern Foundation
“Researchers and practitioners from PAI, Microsoft, OpenAI, Stanford HAI, the Alan Turing Institute, and other leading organizations have come together to call for stronger safeguards for AI agents. Together, we've released Prioritizing Real-Time Failure Detection in AI Agents, a report that argues for proactive steps before these systems scale. While GenAI systems already raise serious concerns, agents directly take actions. By managing sensitive data, coordinating workflows, or eventually negotiating contracts, these systems introduce additional risks that current safeguards cannot fully address. To manage these risks, the report emphasizes the importance of real-time failure detection: automated monitoring systems that can flag anomalies, halt execution, or escalate to human oversight. We also identify gaps in technical methods, evaluation standards, and policy frameworks, and are building on this work to develop industry guidelines for agent monitoring.” -Madhulika Srikumar, Head of AI Safety Governance, Partnership on AI
“As advanced AI systems have moved from the lab into consumer products, they increasingly feature personalized experiences intended to make them more "useful" to users. These features raise many similar concerns as personalized apps and recommender systems that are typical of consumer tech products, but at the time we wrote this report, advocacy around AI had largely remained oriented around privacy issues within model training data and frontier risks like enabling the development of damaging weapons or loss of human control. We wanted to highlight the ways in which the addition of personalized features to interactive AI-powered products ought to be top of mind for advocates who care about digital discrimination, corporate use of sensitive data, and how certain business models incentivize the aggressive collection of data in ways that can lead to widespread and cumulative harm.” -Miranda Bogen, Director of AI Governance Lab, Center for Democracy & Technology
“Giving up valuable face data should not be the price to fly. When TSA announced it would be rolling out facial recognition technology to over 430 airports in the United States, we felt it was important to capture passengers’ lived experiences with TSA FRTs from checkpoints around the country and provide the public with the opportunity to make an informed decision. If facial surveillance technologies become normalized at airports, it invites wider use in other areas of life, such as sporting events, schools, and even hospitals. This program is collecting our most sensitive data. The public should be given the opportunity to choose whether or not to participate while being treated with dignity and respect.” -Dr. Joy Buolamwini, Founder, Algorithmic Justice League
"The 18 organizations that support PAI's Synthetic Media Framework range from AI-developers and social media platforms to news and civil society organizations. In the case studies they shared with us, we saw real-world application of the Framework principles and ways in which guidance needed to be expanded. We also saw the opportunity for action from stakeholders across the board and developed clear recommendations in hopes of advancing towards a future where synthetic media serves creativity and communication while preserving truth, trust, and shared reality." -Claire Leibowicz, Director of AI, Trust, & Society, Partnership on AI
"Strengthening information integrity in elections is no longer just a technical challenge–it’s a democratic imperative. Around the world, voters are navigating unprecedented waves of mis- and disinformation shaped by geopolitical tensions, platform dynamics, and emerging AI capabilities. Our recent 2024-25 partnership with UNDP, alongside my 2025 work on election integrity playbooks and cross-sector panels at All Tech Is Human, has been essential in translating global election-integrity principles into practical tools that governments, civil society, and election authorities can use in real time. Together, these efforts are building the durable guardrails and infrastructure that help safeguard public trust, reinforce democratic resilience, and ensure communities have access to accurate, reliable information during the most consequential civic moments–now and in the years ahead.” -Alexis Crews, Senior Fellow for Information Integrity, All Tech Is Human
"Generative AI models have permeated daily life, and newer versions that accept and generate images are being released in rapid succession. In parallel, IIA - one of the most common forms of online gender-based violence - is growing worldwide. In light of this, it is increasingly important to evaluate the role of generative AI in the creation of IIA and reevaluate the role of various actors in preventing and responding to IIA. This report assesses the effectiveness of existing mitigation and remediation tools on social media and communication platforms, NGO support services and third-party tools, and legislations and policies across different countries." -Dhanya Lakshmi, Consultant, Humane Intelligence
"We highlighted this growing issue via the AI Evaluation Chart Crisis blog post because the pressures that drive misleading evaluations, including speed, competition, and opaque metrics, are undermining the scientific foundations of AI governance. When evaluation charts influence regulation, procurement, and public trust, accuracy becomes a matter of public interest. EvalEval is working to reclaim evaluation as a scientific discipline rather than a marketing function and to create standards that the field can be held accountable to." -Avijit Ghosh, Lead, EvalEval Coalition
“When ARVA and Data & Society began our research in 2023, ‘red-teaming’ was a battleground. Different fields were talking past each another. Some would dismiss GenAI red-teaming as just basic testing. Others worried that public testing was safety theater.
Our Red-Teaming in the Public Interest report argues that the path forward is specificity. Don’t debate red-teaming in the abstract. Investigate the rigorous design choices that make different forms of red-teaming effective in different contexts.
As a technical governance practitioner, I believe no community can do this work on its own. Improving basic testing is vital. So is testing that embraces a critical thinking mindset to challenge normal routines. Downplaying expert testing is not wise. So is failing to experiment with the many ways public testing can serve the public interest.” -Borhane Blili-Hamelin, Officer and Director, AI Risk and Vulnerability Alliance
“The annual CAIDP Index is the most comprehensive and trusted source of comparative AI policy analysis covering 80 countries. More than 1,000 researchers and reviewers contribute to updates and reviews. We want to assess progress, identify emerging trends, and encourage countries to reduce the gap between their commitments and practices. CAIDP Index has influenced many national AI policy developments, and also created a community of advocates.” -Merve Hickok, President, Center for AI and Digital Policy (CAIDP)
"As data center construction intensifies in many regions around the world, it's crucial that we establish better metrics for weighing the environmental and social impacts of these infrastructures in specific locations. Within the US, we have yet to establish standards for measuring the water, energy, and carbon costs of data centers, and we also need to gather qualitative data on quality of life issues that affect neighboring communities, from air quality and other public health matters like noise pollution to land use and higher utility bills. It is important for policymakers, advocates, researchers, and grassroots organizations to work together to gather empirical evidence in the face of financial speculation, the erosion of environmental protections, and tech lobbyist talking points." -Tamara Kneese, Director, Climate, Technology, and Justice Program, Data & Society Research Institute
“AI-powered nonprofits are modeling what responsible AI can look like. Fast Forward's data shows that 71% of AI-powered nonprofits respondents already have processes in place to assess and mitigate AI risks, but 41% point to the lack of in-house technical expertise as a barrier to meeting their ethical ambitions. In grantmaking, funders should encourage transparent policies, support ethics training, and offer governance templates — and provide the funding necessary to do so. By supporting responsible practices, funders signal that equity and accountability are as important as innovation.” -Kevin Barenblat, Co-Founder, Fast Forward
This report was made possible through the support of The Patrick J. McGovern Foundation.