Earlier this year, hundreds of engineers at one of the world’s biggest technology companies did something you do not often see.
They signed a letter urging their employer to reject military work. The company acknowledged the protest, refused to change course and carried on much as before. That small, failed rebellion has stayed with me because of what it reveals about who actually decides how the artificial intelligence (AI) we are increasingly coming to depend on will be used.
Here is what happened, according to reports by Fortune and The Information. In late April, news emerged that the company had signed a classified agreement allowing the military to run its AI on secure networks for, in the contract’s wording, any lawful purpose.
Close to 600 employees reportedly signed an open letter opposing the deal, with the number later rising towards 1,000. The company’s leadership did not yield. An internal memo told staff that the company “proudly” works with the military. Despite the growing number of signatures, the protest achieved nothing.
The issue is not any one company or any one country’s military. It is the question of who gets to draw the line on how AI is used. This year, that question confronted the major AI companies, and they did not all respond in the same way.
Several, according to Fortune, accepted broad “all lawful use” terms with the Pentagon. One major firm refused. The response was blunt: the Pentagon reportedly directed the military and its contractors to stop using that company’s products within six months and designated it a supply-chain risk. The company is now challenging the decision in court.
The company that tried to hold a firmer line is therefore facing pressure, while those that agreed to the terms have continued without comparable consequences. Whatever one’s view of the companies involved, the lesson is clear: saying no can carry a real cost.
This is not merely a Silicon Valley soap opera. Its consequences reach us, and fairly directly.
To put it plainly, we are downstream of other people’s consciences. Malaysia and the region around us do not develop frontier AI models. We buy them, host them in the data centres being built across the country and, before long, may deploy them in our hospitals, banks and security agencies.
The rules governing what these systems may and may not do are being debated – or quietly abandoned – in boardrooms and staff petitions thousands of kilometres away. Whatever decisions are made there, we inherit their consequences. We have little say in the process.
For years, there was a comforting assumption that the people building this technology would act as a brake on its misuse – that engineers with a conscience would refuse to cross certain lines. That assumption has now been tested in public, and it has not held. Nearly 1,000 signatures failed to move management.
Fortune points to why employees have less leverage than they did in 2018, when a similar revolt at the company behind the current employee protest led it to abandon a military AI contract known as Project Maven. Layoffs and an uncertain job market have made employees more cautious about speaking out. Company leaders can also argue that if they refuse the work, a competitor will take it instead. Once every company can make that argument, none has to stand firm.
There is also the question of companies shifting their own boundaries when it suits them. In early 2025, the company involved in the Project Maven controversy quietly removed its public commitment not to develop AI for weapons or surveillance that violated accepted norms. A pledge many had regarded as enduring was simply deleted, with little fanfare. That is worth remembering: a corporate red line is written in pencil, and it lasts only as long as keeping it is cheaper than erasing it.
So where does that leave us? We cannot continue outsourcing our own red lines to foreign engineers and corporate boards. They have shown how quickly those boundaries can give way under pressure. If we are going to deploy this technology across public life, the limits on its use here must be set here, through our own laws, rather than borrowed from whatever a vendor’s values happen to permit this quarter.
A few practical steps follow. When ministries, government agencies or banks procure powerful AI systems, the safeguards that matter to us must be written into the contracts we sign and be enforceable in Malaysian courts. These should include safeguards against machines making autonomous decisions that could harm individuals, blanket surveillance of ordinary citizens and serious decisions being made without an identifiable human being accountable for them.
A company’s published values are a brochure. The contract is the only promise that can be enforced. In matters involving security, a human being must remain in control, without exception. AI can gather, analyse and flag information faster than any officer, and that capability is useful. But it must not be allowed to decide on its own to act against an individual. The consequences of a wrong decision in such circumstances could be fatal. Human oversight is therefore essential.
Underlying all this is the uncomfortable question of leverage. The companies that yielded did so partly because refusing the work seemed too costly. A country without its own AI expertise is in much the same position: it has little choice but to accept the terms offered because it lacks the capability to walk away.
Every Malaysian engineer we train and every capability we develop locally strengthens our ability to set our own conditions rather than accept someone else’s.
The hard truth is that the most powerful technology of our age is being shaped more by what pays than by what is right. Those inside the industry who tried to object have discovered just how little their objections count. That is no reason to throw up our hands. It is a reason to stop assuming that somebody else, somewhere else, will do our thinking and make our refusals for us.
The engineers who signed that letter were ultimately overruled by their employer. The least we can do is heed the warning their protest failed to deliver and draw our own line in law before this technology becomes too deeply embedded in our lives to be constrained. Trying to do so afterwards may be too late.
The views expressed here are the personal opinion of the writer and do not represent those of Twentytwo13.