We Should Ensure Accountablity In The AI Military Sector – OpEd
In September 2025, Microsoft terminated some of its services to the Israeli military following an investigation that suggested Israel is using the technology for mass surveillance of Palestinians. Microsoft then informed the Israel Defense Ministry of suspending specific cloud storage and AI services and technologies.
However, this is unfortunately not usually the case; AI companies have consistently shown that they do not adhere to the pledges they made and are willing to violate their policies that prohibit them from selling their models to military entities. For instance, in December 2024, a news report exposed that OpenAI is working with Anduril to supply the U.S. military with AI. Meta has also opened its model for military use in the U.S. Alarmingly, companies such as Google and OpenAI have quietly altered their usage policies to permit military purposes, even worse, in some cases, such as project Nimbus, the terms of the amended contract have been kept confidential.
Far worse, in the AI military sector, we are facing a significant legal gap because in many of its areas, there is no dedicated treaty regulating the use of this technology. In fact, countries such as the U.S. and the U.K. are hindering the effort to reach an international consensus on banning killer robots.
In the absence of this legal protection, society can still play a crucial role in safeguarding itself.
The advancement in AI, namely the introduction of large language models, has brought with it a huge promise to the military sector. AI models are currently leveraged in different areas on the battlefield. In several recent wars, AI models have been documented as being used in mass surveillance to intercept thousands of communications, as well as in the creation and dissemination of fabricated content to disinform the enemy about the unfolding events surrounding the war. In battlefields that took place in countries such as Libya, Azerbaijan, and Ukraine, lethal autonomous systems with no human intervention were used, according to verified reports.
There is a high cost associated with the error from the nature of automated systems which may rely on faulty data or biased decisions. More importantly, there is concern about these systems falling into the hands of authoritarian and repressive regimes, such as Russia and China, where we lack transparency and certainty regarding their applications in the field. Even inside democratic countries such as the U.S., the danger still exists. Claudi’s recent move to partner with Palnatire, which includes processing secret government data, was criticized by some as a deviation from its declared safety aim.
In light of this potential risk and the current legal shortcomings, we have the moral duty to work in four directions to deter AI companies and governments, for that matter, from pursuing their interests at the expense of endangering the safety of citizens.
Media pressure and coverage are powerful tools for pushing back against AI companies and governments. The last protest in front of OpenAI, demanding the cessation of military partnership, and the petition by Google and Amazon workers to condemn the Nibmus project were covered by leading news websites. Similar pressure led Google in 2018 to not renew its contract with the Defense Department. Documentary films like the Slaugthersbots and Unknown: Killer Robots are helping raise awareness about the issue. Moreover, the open letters by known figures in the field also put pressure.
The efforts by watchdog and rights groups such as No Tech for apartheid, Human Rights Watch, and Stop Killer Robots are crucial in exposing the misuse of AI in the military. The Midas project lets the public know when AI companies discreetly change their polices without any announcement.
There is also a dire need to guard whistleblowers who put their jobs and perhaps even their existence on the line to tell the truth. In April 2024, six Israeli intelligence officers shared with an investigative journalist how the army is using an advanced automated system called Lavender to target. In another case, a Google engineer resigned in protest over being sent to a military drone project; a legal framework is required to protect them against retaliation and encourage others to speak out in the future.
Divestment from AI companies could also be effective. Although achieving the goals of these campaigns is practically difficult, given that the products of AI companies are widely adopted by society and are essential for daily life, they remain useful in exposing the companies to the public and acting as a form of solidarity.
The legal gap will likely persist for a while, as there is no guarantee that the UN will succeed in regulating this technology by 2025, as it plans. AI companies and governments will continue to exploit it for their benefit. However, we should keep experimenting with different tactics and strategies to lessen the harm.
Like what you read?
Please consider supporting Eurasia Review. Thank you for your consideration!
