Article
AI
Creed
Ethics
5 min read

Whistleblowing: what if your CEO is a Caesar?

What are the boundaries of legitimate protest?

Professor Charles Foster is a Fellow of Exeter College, Oxford, and a member of the Oxford Law Faculty.

On a conference stage, a seated speaker leans back and opines
Sam Altman, CEO of OpenAI.
TechCrunch, CC BY 2.0, via Wikimedia Commons .

If you discovered that the company you worked for was doing work that posed an existential risk to humanity, would you consider yourself entitled – or perhaps morally obliged – to blow the whistle? 

This issue provoked a recent open letter from current and former employees at AI companies including Sam Altman’s OpenAI, asserting that the laws protecting whistleblowers are inadequate because they typically focus on illegal activity – and the AI companies concerned are doing nothing which is (yet) illegal. It called for companies to take a number of steps (including not entering into or enforcing agreements prohibiting the raising of risk-related concerns). 

Some might say that if an employee takes the company’s money, that money should buy loyal silence, and that if the public interest demands a different approach, the remedy is the extension to risk-related concerns of existing whistleblower legislation. But unless and until that legislation is extended, should we applaud conscience-driven breaches of contract?  

What about breaches of the criminal law for morally justifiable reasons – for instance to draw attention to the risks that the protestors say are associated with climate change?  

The reality of modern corporate governance means that the CEO may be more practically Caesarean than a country’s government. 

Christian debate about these issues has traditionally turned on two Bible texts. Paul, in writing to those in a Roman church, declares: ‘Let every person be subject to the governing authorities, for there is no authority except from God, and those authorities that exist have been instituted by God. Therefore whoever resists authority resists what God has appointed….the authority… is the agent of God.’ And Jesus, in Matthew's gospel, advises us to ‘render to Caesar that which is Caesar’s, and to God that which is God’s.’ 

Who are the ‘authorities’ spoken of by Paul? Who is the modern Caesar spoken of by Jesus? Presumably in each case – in a parliamentary democracy – it is the combined legislature and executive of the day. Perhaps, these days, we should translate ‘Caesar’ as ‘the social contract’. But does this mean that (if we take these injunctions seriously) we should regard ourselves as bound not to commit criminal offences (which are offences against the state), but should feel no corresponding inhibition about breaching private law obligations, such as those owed under contracts of employment? My instinct is to say that this is indeed what it means, but that is not self-evident. After all, much employment law is statutory – an emanation of Parliament, and the reality of modern corporate governance means that the CEO may be more practically Caesarean than a country’s government. 

Rendering the right thing to Caesar in a theocracy such as Byzantium might mean something very different in a modern tyranny or a democracy.

Should Christians, though, feel constrained by these scriptural passages? Both Paul and Jesus seemed to think that there was little point in establishing lasting social, legal or governmental structures because the end times were just around the corner. Jesus thought that some of his audience would still be alive when the Son of Man returned to complete the messianic project without any help from any secular governor. Paul’s belief that the Second Coming of Christ was at hand was behind his advice that the unmarried (unless they really couldn’t stay celibate) should remain unmarried and get on with the urgent business of preparing for the imminent in-rush of the true Kingdom. Both Jesus and Paul were dramatically wrong about the chronology. Why, then, should we take seriously advice about the regulation of society that was based on their mistake? Should Paul’s advice to those Romans be read as pragmatism – intended by him to convince rulers that Christians wouldn’t make trouble, and that therefore the Christians should be left alone? He may have thought that a shabby compromise with secular powers didn’t matter much because it wouldn’t last long.  

Even if these texts are in some meaningful sense authoritative, what do they mean for modern life? As ever, the devil (and potentially the angel) is in the detail, and Paul and Jesus left the church to work out the relevant details. There is no consensus. Rendering the right thing to Caesar in a theocracy such as Byzantium might mean something very different in a modern tyranny or a democracy. Only in a few situations is the correct answer obvious: no one would doubt that those martyred for refusing to worship the Caesar of the day had made the (or at least a) right choice. But as soon as we move away from such cases the waters get muddy. Would Paul have denounced Dietrich Bonhoeffer for the plot to kill Hitler? If so, would he have been right? It cannot be seriously argued that it is illegitimate to protest against the policies of the day, any more than it could be suggested that Paul requires us to cast our vote in favour of the currently ruling party. 

What, then, are the boundaries of legitimate protest?  

Suppose that AI really does pose a threat to the whole of humanity. Does ‘rendering to God’ not then demand, in a private law context, that the whistle be blown, even if it involves a breach of a contractual obligation? It seems at least arguable.  

Is a breach of the criminal law – for instance in the case of climate change protestors – different? It may well be.  

In England the law has evolved a nuanced approach to ethically motivated criminality. That approach was recently displayed in the sentencing of five Extinction Rebellion activists for criminal damage to the premises of a bank. The judge accepted that each defendant believed that the bank was culpably involved in funding fossil fuel extraction projects, and that such projects endangered the planet. He noted that Lord Hoffman had said: ‘People who break the law to affirm their belief in the injustice of a law or government action are sometimes vindicated by history [for instance the suffragettes]. It is the mark of a civilized community that it can accommodate protests and demonstrations of this kind. But there are conventions which are generally accepted by the law-breakers on one side and the law-enforcers on the other. The protestors behave with a sense of proportion and do not cause excessive damage or inconvenience. And they vouch the sincerity of their beliefs by accepting the penalties imposed by the law.’ In return, he went on, the state behaves with restraint, and the judiciary imposes sentences which take the conscientious motives into account. 

This approach, said the sentencing judge, amounts to a ‘social compact between the courts and protestors.’  

Perhaps, in the realm of the criminal law, that sort of social compact encodes the relevant moral and theological principles as well as anything can.  

Article
AI
Attention
Culture
5 min read

Will AI’s attentions amplify or suffocate us?

Keeping attention on the right things has always been a problem.

Mark is a research mathematician who writes on ethics, human identity and the nature of intelligence.

A cute-looking robot with big eyes stares up at the viewer.
Robots - always cuter than AI.
Alex Knight on Unsplash.

Taking inspiration from human attention has made AI vastly more powerful. Can this focus our minds on why attention really matters? 

Artificial intelligence has been developing at a dizzying rate. Chatbots like ChatGPT and Copilot can automate everyday tasks and can effortlessly summarise information. Photorealistic images and videos can be generated from a couple of words and medical AI promises to revolutionise both drug discovery and healthcare. The technology (or at least the hype around it) gives an impression of boundless acceleration. 

So far, 2025 has been the year AI has become a real big-ticket political item. The new Trump administration has promised half a trillion dollars for AI infrastructure and UK prime minister Keir Starmer plans to ‘turbocharge’ AI in the UK. Predictions of our future with this new technology range from doom-laden apocalypse to techno-utopian superabundance. The only certainty is that it will lead to dramatic personal and social change. 

This technological impact feels even more dramatic given the relative simplicity of its components. Huge volumes of text, image and videos are converted into vast arrays of numbers. These grids are then pushed through repeated processes of addition, multiplication and comparison. As more data is fed into this process, the numbers (or weights) in the system are updated and the AI ‘learns’ from the data. With enough data, meaningful relationships between words are internalised and the model becomes capable of generating useful answers to questions. 

So why have these algorithms become so much more powerful over the past few years? One major driver has been to take inspiration from human attention. An ‘attention mechanism’ allows very distant parts of texts or images to be associated together. This means that when processing a passage of conversation in a novel, the system is able to take cues on the mood of the characters from earlier in the chapter. This ability to attend to the broader context of the text has allowed the success of the current wave of ‘large language models’ or ‘generative AI’. In fact, these models with the technical name ‘Transformer’ were developed by removing other features and concentrating only on the attention mechanisms. This was first published in the memorably named ‘Attention is All You Need’ paper written by scientists working at Google in 2017. 

If you’re wondering whether this machine replication of human attention has much to do with the real thing, you might be right to be sceptical. That said, this attention-imitating technology has profound effects on how we attend to the world. On the one hand, it has shown the ability to focus and amplify our attention, but on the other, to distract and suffocate it. 

Attention is a moral act, directed towards care for others.

A radiologist acts with professional care for her patients. Armed with a lifetime of knowledge and expertise, she diligently checks scans for evidence of malignant tumours. Using new AI tools can amplify her expertise and attention. These can automatically detect suspicious patterns in the image including very fine detail that a human eye could miss. These additional pairs of eyes can free her professional attention to other aspects of the scan or other aspects of the job. 

Meanwhile, a government acts with obligations to keep its spending down. It decides to automate welfare claim handling using a “state of the art” AI system. The system flags more claimants as being overpaid than the human employees used to. The politicians and senior bureaucrats congratulate themselves on the system’s efficiency and they resolve to extend it to other types of payments. Meanwhile, hundreds of thousands are being forced to pay non-existent debts. With echoes of the British Post Office Horizon Scandal, the 2017-2020 the Australian Robo-debt scandal was due to flaws in the algorithm used to calculate the debts. To have a properly functioning welfare safety net, there needs to be public scrutiny, and a misplaced deference to machines and algorithms suffocated the attention that was needed.   

These examples illustrate the interplay between AI and our attention, but they also show that human attention has a broader meaning than just being the efficient channelling of information. In both cases, attention is a moral act, directed towards care for others. There are many other ways algorithms interact with our attention – how social media is optimised to keep us scrolling, how chatbots are being touted as a solution to loneliness among the elderly, but also how translation apps help break language barriers. 

Algorithms are not the first thing to get in the way of our attention, and keeping our attention on the right things has always been a problem. One of the best stories about attention and noticing other people is Jesus’ parable of the Good Samaritan. A man lies badly beaten on the side of the road after a robbery. Several respectable people walk past without attending to the man. A stranger stops. His people and the injured man’s people are bitter enemies. Despite this, he generously attends to the wounded stranger. He risks the danger of stopping – perhaps the injured man will attack him? He then tends the man’s wounds and uses his money to pay for an indefinite stay in a hotel. 

This is the true model of attention. Risky, loving “noticing” which is action as much as intellect. A model of attention better than even the best neuroscientist or programmer could come up with, one modelled by God himself. In this story, the stranger, the Good Samaritan, is Jesus, and we all sit wounded and in need of attention. 

But not only this, we are born to imitate the Good Samaritan’s attention to others. Just as we can receive God’s love, we can also attend to the needs of others. This mirrors our relationship to artificial intelligence, just as our AI toys are conduits of our attention, we can be conduits of God’s perfect loving attention. This is what our attention is really for, and if we remember this while being prudent about the dangers of technology, then we might succeed in elevating our attention-inspired tools to make AI an amplifier of real attention. 

Join with us - Behind the Seen

Seen & Unseen is free for everyone and is made possible through the generosity of our amazing community of supporters.

If you’re enjoying Seen & Unseen, would you consider making a gift towards our work?

Alongside other benefits (book discounts etc.), you’ll receive an extra fortnightly email from me sharing what I’m reading and my reflections on the ideas that are shaping our times.

Graham Tomlin

Editor-in-Chief