Shaping the future of ethical AI

In 2024, Google’s Generative AI Gemini became the target of nation-state hackers attempting to bypass safeguards and steal sensitive data. The attackers failed but only just.[1] Incidents like this highlight how fast AI risks are evolving and why responsible governance cannot wait.

At Sarasin, we see the development and deployment of artificial intelligence (AI) as one of the defining challenges of our time. These technologies offer transformative potential but also come with serious risks. From algorithmic bias and misinformation to cybercrime and loss of privacy, the implications for individuals, societies and markets are profound.

That is why ethical AI has been one of our top stewardship priorities since 2022. We believe investors have a critical role to play in ensuring that AI is developed and used responsibly. Through active engagement, voting and coalition-building, we are pressing companies to improve transparency, adopt stronger safeguards and protect long-term value.

Why AI governance matters

AI technology is neutral. But the use of it is not. Whether powering search engines, social media feeds or enterprise platforms, AI systems reflect the assumptions and priorities of those who build and train them. Without proper oversight, they can entrench discrimination, allow the spread of false information and expose people and institutions to fraud, surveillance or harm.

The risks are no longer theoretical. Social media platform X (formerly Twitter) saw its advertising revenues fall by more than 50% after relaxing its content moderation policies – a shift that coincided with a spike in hate speech and disinformation.[2] For investors, the message is clear: when AI risks are not properly managed, financial consequences often follow.

Despite impressive progress in recent years, largely resulting from investor pressure, in many companies, the systems for managing those risks remain vulnerable. Ethical principles are often published without clear implementation strategies or board-level oversight. Security testing, when it happens, does not always reveal all the results, and human rights impact assessments (HRIA) are rare. Our goal is to change that by helping companies translate high-level commitments into meaningful action.

A growing global response

There is growing recognition that AI development and deployment need firmer guardrails. In September 2024, the United Nations adopted the Global Digital Compact, which is the first international framework aimed at promoting a safe and inclusive digital environment. Among its objectives are protecting human rights, ensuring digital inclusion and mitigating the risks associated with AI technologies.

At the investor level, momentum is also building. In February 2024, the World Benchmarking Alliance (WBA) published an updated Investor Statement on Ethical AI, signed by 48 institutions representing $8.5 trillion in assets under management.[2] Sarasin was a contributor to this work, calling for clearer disclosure of the tools companies use to operationalise their AI principles, including safety testing, human oversight and human rights due diligence.

To support our engagements, we developed a proprietary matrix of AI-related risks and mitigation tools tailored to specific business models. For developers of generative AI and large language models, we recommend measures such as red teaming (a cybersecurity practice where ethical hackers, known as the red team, simulate real-world cyberattacks to test an organisation's security defences), watermarking, output analysis including HRIA and reinforcement learning from human feedback. These tools provide practical ways to reduce risk and enhance trust.

Engagement in action

We have taken the lead on AI-related engagements with several high-profile companies through our stewardship programme and our involvement in the WBA’s Collective Impact Coalition on Ethical AI. Our focus is to ensure that governance systems and resulting guardrails keep pace with the rapid evolution of AI.

 Alphabet: confronting cyber threats and disinformation

Alphabet is the parent company of Google and YouTube, and widely viewed as a front-runner in AI development. It has established ethics councils, published responsible AI frameworks, and co-founded the Frontier Model Forum to promote safe practices.

Yet significant risks remain. In early 2024, we raised concerns about the malicious use of Alphabet’s Gemini model. Shortly afterwards, Google published a report showing how actors linked to China, Russia, Iran and North Korea had attempted to exploit Gemini for phishing, cybercrime and disinformation.[3]

While many of Google’s safeguards held up, the incident underlined the importance of ongoing risk assessment. We have since called for more systematic use of HRIAs and stronger oversight of YouTube’s content moderation practices.

As lead investors in Alphabet’s engagement through the WBA, we coordinated a collective letter to senior executives, voted against multiple directors at the 2024 AGM, and supported AI-related shareholder proposals calling for improved governance and transparency. Alphabet responded with links to resources, including updates on its AI principles, privacy practices and red teaming protocols. It is a welcome step, though several concerns remain. We continue our outreach for collective engagement with the company.

Meta: pushing for accountability amid policy shifts

While Alphabet has shown a degree of transparency, our experience with Meta illustrates how progress can quickly unravel. In September 2024, we participated in a joint investor call with Meta’s governance team; we then arranged a letter, backed by $3.6 trillion in assets under management, urging greater clarity on content moderation, child safety, privacy and data training practices.[4]

Meta initially responded constructively. But in early 2025, the tone shifted. Following leadership changes at board level, the company dismantled its fact-checking programme in the US and replaced it with a crowdsourced moderation tool known as Community Notes.[5]

Community Notes has a mixed record. On the X platform, where the tool was first introduced, independent analysis found that 74% of misleading posts remained uncorrected, and when notes were added, false content continued to receive far more views than corrections.[6]

We are now evaluating whether Meta’s changes represent a rollback of earlier reforms. The lack of formal HRIAs and reduced transparency around data use have heightened our concerns. We are following up to assess whether the company remains committed to responsible AI governance, and will continue to advocate for stronger protections against disinformation, bias and privacy breaches.

In May 2025, we pre-declared our votes at the Meta AGM on our website: we voted against five of its directors on accountability concerns, the executive remuneration and the auditor, we supported one-year Say on Pay frequency and eight shareholder resolutions, including five asking for more efforts towards ethical AI.[7]

ServiceNow: constructive dialogue and positive steps

In contrast to the challenges we have encountered with some consumer-facing platforms, our engagement with enterprise software provider ServiceNow has been constructive and encouraging.

As lead investors in the WBA’s engagement, we raised concerns about how ServiceNow’s generative AI tools could be used in ways that introduce bias or lack transparency. Following our outreach, the company published its Human-Centred Responsible AI Guidelines and shared details of its governance approach.

ServiceNow’s model emphasises shared responsibility, with mechanisms for customers to participate in AI risk assessment and mitigation. These include detailed model cards, human oversight systems, differentiated access controls and risk evaluation tools. During our engagement call, we also discussed the company’s collaboration with academic institutions and standards bodies such as the US National Institute of Standards & Technology.

We recommended greater clarity on board-level governance and encouraged the company to share real-world examples and independent evaluations. ServiceNow was receptive to our feedback and committed to further improvements. We will continue to monitor progress and support its efforts to embed responsible AI practices.

How we exert influence

Our stewardship work combines direct dialogue with the board or senior company leadership, coordinated action through investor coalitions, and the strategic use of voting rights. We also contribute to public consultations and frameworks that shape global norms.

Through these levers, we aim to move companies from words to action – from publishing AI principles to implementing meaningful safeguards that stand up to scrutiny.

AI continues to evolve rapidly. One emerging focus for our engagement is the rise of autonomous AI agents – systems capable of making decisions and acting independently. These technologies promise new efficiencies, but also introduce major risks. Can such agents be trusted? Who is accountable if they cause harm? Could they be hijacked or manipulated?

We will be pressing companies to put in place robust guardrails, including anti-hijack safeguards, transparency requirements and clearer lines of accountability.

AI is already shaping the way we work, communicate and interact with the world. The decisions made today will determine whether these technologies enhance human wellbeing or undermine it.

At Sarasin, we believe investors have both the responsibility and the influence to shape a safer AI future. Through persistent engagement, coalition-building and public advocacy, we will continue to push for ethical AI, ensuring technology serves society, not the other way around.

[1] https://cloud.google.com/blog/topics/threat-intelligence/adversarial-misuse-generative-ai

[2] www.worldbenchmarkingalliance.org/impact/investor-statementfor-ethical-ai-2024/

[3] www.reuters.com/technology/us-ad-revenue-musks-x-declinedeach-month-since-takeover-data-2023-10-04/

[4] https://sarasinandpartners.com/stewardship-post/metaplatforms-pre-declaration-of-votes/

[5] https://about.fb.com/news/2025/01/meta-morespeech-fewer-mistakes/

[6] https://counterhate.com/research/rated-not-helpful-xcommunity-notes/

[7] https://sarasinandpartners.com/stewardship-post/metaplatforms-pre-declaration-of-votes/

Important information

This document is intended for retail investors and/or private clients.  You should not act or rely on this document but you should contact your professional adviser.

This document has been issued by Sarasin & Partners LLP of Juxon House, 100 St Paul’s Churchyard, London, EC4M 8BU, a limited liability partnership registered in England and Wales with registered number OC329859, and which is authorised and regulated by the Financial Conduct Authority with firm reference number 475111. This document has been prepared for informational purposes only and is not a solicitation, or an offer to buy or sell any security. The information on which the material is based has been obtained in good faith, from sources that we believe to be reliable, but we have not independently verified such information and we make no representation or warranty, express or implied, as to its accuracy. All expressions of opinion are subject to change without notice. This document should not be relied on for accounting, legal or tax advice, or investment recommendations. Reliance should not be placed on the views and information in this material when taking individual investment and/or strategic decisions.

The value of investments and any income derived from them can fall as well as rise and investors may not get back the amount originally invested. If investing in foreign currencies, the return in the investor’s reference currency may increase or decrease as a result of currency fluctuations. Past performance is not a reliable indicator of future results and may not be repeated. Forecasts are not a reliable indicator of future performance.

Neither Sarasin & Partners LLP nor any other member of the J. Safra Sarasin Holding Ltd group accepts any liability or responsibility whatsoever for any consequential loss of any kind arising out of the use of this document or any part of its contents. The use of this document should not be regarded as a substitute for the exercise by the recipient of their own judgement. Sarasin & Partners LLP and/or any person connected with it may act upon or make use of the material referred to herein and/or any of the information upon which it is based, prior to publication of this document. Where the data in this document comes partially from third-party sources the accuracy, completeness or correctness of the information contained in this publication is not guaranteed, and third-party data is provided without any warranties of any kind. Sarasin & Partners LLP shall have no liability in connection with third-party data. © 2025 Sarasin & Partners LLP – all rights reserved. This document can only be distributed or reproduced with permission from Sarasin & Partners LLP. Please contact [email protected].