Military use, cyberattacks, political messaging: How governments are approaching AI

Artificial intelligence has become a major part of nearly every aspect of daily life in recent years. Yet alongside its benefits, AI also carries significant risks, ranging from minor errors to potentially life-threatening failures and even scenarios that could bring countries close to war.

Governments have taken markedly different approaches to the issue. Some are seeking to encourage citizens, companies and state institutions to adopt AI more broadly, while others are emphasizing the risks and imposing tighter restrictions.

One of the recurring themes of US President Donald Trump’s current term has been the need to preserve the United States’ position as the world’s technological leader, above all in its competition with China. Trump himself has repeatedly expressed fascination with artificial intelligence and described it as a critical tool for maintaining American power.

As part of that approach, Trump has moved to ease regulations in order to accelerate the adoption of AI models and their integration into government bodies and other institutions.

One prominent example is the US military, which used several AI models, including specialized versions, during the war with Iran. Their reported uses included processing and analyzing large volumes of intelligence and helping select targets for strikes.

That made the sweeping restrictions imposed by the Trump administration in June 2026 on the most advanced models developed by Anthropic and OpenAI particularly striking. The move marked a significant shift in Washington’s approach to AI, from regulatory easing toward far tighter restrictions.

The reason was an alarming leap in the models’ cyber capabilities and fears that actors such as China, Russia or hostile hackers could use them to disrupt or even cripple critical national infrastructure.

The companies said they had introduced strong safeguards intended to prevent the models from identifying, analyzing and exploiting security vulnerabilities, including restrictions on generating code capable of attacking those weaknesses.

US intelligence agencies, however, argued and demonstrated that those safeguards could be circumvented. That finding prompted wider government involvement and ultimately led to the stricter controls.

At first glance, the response might appear excessive, particularly because similar concerns have been raised before. But major cyberattacks in recent years, including in the United States, have caused extensive economic damage and disrupted critical infrastructure.

In previous cases, cybersecurity specialists were eventually able to contain the attacks within days. Malware generated by artificial intelligence rather than by a human operator, however, could prove far more difficult to confront.

Under pressure from the administration, Anthropic and OpenAI ultimately complied with the White House demands. By the end of June 2026, the rollout of the models had been halted, and they are now available only to a limited group of US companies that have received prior government approval.

Access is restricted to authorized American organizations working to protect critical infrastructure.

Like Washington, Beijing has developed a complicated relationship with artificial intelligence and the models now available to the public.

Competition with the United States is also shaping Chinese policy. In recent months, reports have suggested that Beijing is considering restricting travel by AI experts over fears that they could defect to the United States, while also debating whether its most advanced models should be made inaccessible outside China.

The Chinese government views artificial intelligence as a strategic tool that could give the country greater power across multiple fields and eventually allow it to overtake the United States.

At the same time, as a highly centralized political system, Beijing is deeply concerned about losing control over the flow of information and public opinion. That has led China to establish one of the world’s fastest-developing and most stringent regulatory systems for AI.

One aspect of that system is the protection of the Communist Party and its ideology. While Western governments tend to focus on issues such as copyright, privacy and job losses, China’s official rules require AI-generated content to “reflect the core socialist values.”

In practice, this means AI systems are not permitted to generate content critical of the Communist Party leadership or President Xi Jinping, or to present alternative interpretations of historical events such as the Tiananmen Square crackdown.

Companies are also required to train their models on “screened and approved” information in an effort to prevent them from absorbing Western democratic ideas.

Another component of China’s regulatory framework is mandatory licensing and legal responsibility for developers.

Unlike in the United States, companies cannot simply release a model and wait to see what happens. The Chinese system works in reverse: algorithms must receive advance approval, be registered in a government database and undergo stringent safety testing before a model can be made available to the public.

In addition, if a model produces an incorrect answer, or one that violates government requirements, responsibility falls on the company that developed it, including potential legal and criminal liability.

The result is that Chinese companies impose extensive self-censorship and strict limitations on their AI systems.

The regulatory framework does, however, include measures that have put China at the forefront of efforts to combat deepfakes and protect consumers.

Beijing was among the first governments to introduce comprehensive rules targeting tools used to create deepfakes. Images and videos generated by AI must also carry labels indicating that they were artificially produced.

There are also broad restrictions on using such technologies to manipulate or imitate a person’s voice or appearance without their explicit consent, part of a wider global effort to combat identity theft and online fraud.

At the same time, these policies pose a significant challenge for China’s technology industry.

The path toward building smarter, more creative and more innovative AI models generally involves training them on enormous quantities of information and allowing them to make unexpected connections.

China’s regulations limit some of that freedom, and some experts believe this is one of the main reasons the United States currently retains an advantage over China despite Beijing’s formidable engineering capabilities.

Despite China’s sometimes rigid approach, that balance may soon begin to shift following the unveiling of a new Chinese AI model called Kimi K3.

Developed by Moonshot, the company says K3 is its most capable artificial intelligence model to date, while several assessments suggest that it performs at a level comparable to leading American models.

And what about Israel?

Despite Israel’s reputation as the “Start-Up Nation,” which gives it a distinctive position in the global AI race, the country presents a mixed picture: it is among the leaders in several areas, while lagging behind in others.

The Israeli government has broadly adopted an approach similar to that of the United States, encouraging rapid growth, seeking to ease burdens on the technology industry and avoiding legislation that could inhibit innovation.

Artificial intelligence is viewed in Jerusalem as a critical engine of both economic growth and military power.

That has led the government to favor public funding for infrastructure, relatively light regulation and broad freedom of action for the private sector, so long as national security is not seen as being immediately at risk.

As part of this policy, Israel has decided to invest heavily in domestic supercomputing infrastructure in an effort to reduce dependence on US-based servers.

At the same time, the Innovation Ministry, the Israel Innovation Authority and MAFAT, the Defense Ministry’s Directorate of Defense Research and Development, are promoting the development of a language model tailored specifically to Israel, with the aim of enabling its secure use in government ministries and across the public sector.