At the end of 2023, Israel aimed to assassinate Ibrahim Biari, a high -level commander in the northern strip of Gaza who had contributed to planning the massacres of 7 October. But the Israeli intelligence was unable to find Mr. Biari, who believed it was hidden in the tunnel network under Gaza.
So the Israeli officers turned to a new military technology infused with artificial intelligence, three Israeli and American officials informed the events. The technology had been developed a decade before but had not been used in battle. Finding Mr. Biari provided a new incentive to improve the tool, therefore the engineers in the 8200 unit of Israel, equivalent of the country of the National Security Agency, soon integrated the IA in it, said people.
Shortly thereafter, Israel listened to the calls of Mr. Biari and tested the Audio Ai tool, who gave an approximate position for where he was making his calls. Using this information, Israel ordered air attacks to hit the area on October 31, 2023, killing Mr. Biari. More than 125 civilians also died in the attack, according to Airwars, a monitor of conflicts based in London.
The audio tool was only an example of how Israel used the war in Gaza to test quickly and deploy military technologies supported by artificial intelligence in a degree that had not been seen before, according to interviews with nine American and Israeli defense officials, who spoke on condition of anonymity because the work is confidential.
In the last 18 months, Israel has also combined the IA with facial recognition software to combine partially obscured or injured faces with real identities, he turned to the AI to fill in potential air attack objectives and created a model of AI in Arabic language to feed a chatbot that could scan and analyze text messages, social media text messages and other Arabic language data, two people with the knowledge of the programs.
Many of these efforts have been a partnership between soldiers enrolled in the 8200 unit and soldiers reserves that work in technological companies such as Google, Microsoft and Meta, three people with knowledge of the technologies have said. The 8200 unit created what has become known as “The Studio”, an innovation hub and a place to combine experts with artificial intelligence projects, said people.
Yet, even if Israel ran to develop the arsenal AI, the deployment of technologies has sometimes led to identifications and incorrect arrests, as well as to civilian deaths, the Israeli and American officials said. Some officials have fought with the ethical implications of the AI tools, which could lead to greater surveillance and other civil murders.
No other nation has been active as Israel in the experimentation of artificial intelligence tools in real -time battles, he said officials of the European and American defense, giving a preview of how these technologies can be used in future wars and how they could also go wrong.
“The urgent need to cope with an accelerated innovation in crisis, largely fed by the AI”, said Hadas Lorber, head of the Institute for Research applied at the Head of the Institute of Technology of Holon Israeli and a former Senior Director at the Israeli National Security Council. “He led to technologies that change the game on the battlefield and the advantages that proved to be fundamental in combat”.
But the technologies “also raise serious ethical issues,” said Mrs. Lorber. He warned that artificial intelligence needs checks and balances, adding that humans should make final decisions.
A spokesman for the Israel army said he could not comment on specific technologies because of their “confidential nature”. Israel “is engaged in the legitimate use and responsible for data technology tools,” he added, adding that the military were investigating the strike to Mr. Biari and was unable to provide further information until the investigation is completed “.
Meta and Microsoft refused to comment. Google said she had “employees who do duty to various countries all over the world. The work that those employees do as reservists are not connected to Google”.
Israel has previously used conflicts in Gaza and Lebanon to experiment and advance technological tools for its military, such as drones, telephone hacking tools and Iron Dome's defense system, which can help intercepts short -range ballistic missiles.
After Hamas launched cross -border attacks in Israel on October 7, 2023, killing more than 1,200 people and taking 250 hostages, AI technologies were quickly authorized for the deployment, they said four Israeli officials. This led to the cooperation between unit 8200 and reserves soldiers in “The Studio” to quickly develop new artificial intelligence skills, they said.
Avi Hasson, CEO of Startup Nation Central, an Israeli non -profit organization that connects investors with companies, said that the reservations of Meta, Google and Microsoft have become crucial to guide innovation in drones and in the integration of data.
“The reservists brought the know-how and access to key technologies that were not available in the army,” he said.
Israel's army soon used the IA to improve its drone fleet. Aviv Shapira, founder and managing director of Xtend, a software and drones company that works with the Israeli military, said that algorithms based on artificial intelligence have been used to build drones to block and keep track of the objectives from afar.
“In the past, Homing's abilities were based on zero on an image of the target,” he said. “Now the IA can recognize and keep track of the object itself – it can be a moving car or a person – with fatal precision.”
Shapira said that its main customers, the Israeli military and the Department of Defense of the United States, were aware of the ethical implications of the AI in the war and discussed the responsible use of technology.
A tool developed by “The Studio” was a model of artificial intelligence in Arabic language known as large linguistic model, three Israeli officers said that are familiar with the program. (The large linguistic model was previously reported by Plus 972, a site of Israelo-Palestinian news.)
Previously, the developers had struggled to create such a model due to the lack of data in Arabic to form technology. When these data were available, it was mainly in Arabic standard written, which is more formal than the dozen dialects used in the spoken Arabic.
The Israeli army did not have this problem, the three officers said. The country was decades of intercepted text messages, transcribed and post -scraped phone calls from social media in the spoken Arab dialects. So the Israeli officers created the large linguistic model in the first months of the war and built a chatbot to perform questions in Arabic. They combined the tool with multimedia databases, allowing analysts to carry out complex searches between images and videos, said four Israeli officials.
When Israel assassinated the leader of Hezbollah Hassan Nasrallah in September, the chatbot analyzed the answers across the world of Arabic language, said three Israeli officers. The technology is differentiated between the different dialects in Lebanon to evaluate the public reaction, helping Israel to evaluate whether there was a public pressure for a counter -member.
Sometimes, the chatbot was unable to identify some modern terms and gear words that have been transitted from English to Arabic, said two officers. This requested Israeli intelligence officials with skills in different dialects to review and correct their work, said one of the officers.
The chatbot sometimes also provided wrong responses – for example, restoring photos of pipes instead of pistols – said two Israeli intelligence officers. Even so, the AI tool significantly accelerated research and analysis, they said.
At the temporary control points set up between the northern and southern Gaza Strip, Israel also began to equip the cameras after the attacks of October 7 with the possibility of scanning and sending high resolution images of the Palestinians to a facial recognition program supported by artificial intelligence.
Even this system sometimes had difficulty identifying people whose faces were obscured. This led to Palestinian arrests and interrogations that were erroneously reported by the facial recognition system, said two Israeli intelligence officers.
Israel has also used artificial intelligence to sift the data accumulated by the intelligence officials on the members of Hamas. Before the war, Israel built an automatic learning algorithm-“lavender” appointed in code-that could quickly order data to hunt the low-level militants. He was trained on a database of confirmed members of Hamas and intended to predict who else could be part of the group. Although the forecasts of the system were imperfect, Israel used it at the beginning of the war in Gaza to help choose the attack goals.
Few objectives loomed bigger than research and eliminating Hamas's senior leadership. Near the top of the list was Mr. Biari, the Hamas commander, that Israeli officials believed played a central role in planning the attacks of 7 October.
The military intelligence of Israel quickly intercepted the calls of Mr. Biari with other members of Hamas but was unable to identify his position. So they turned to the audio tool supported by the AI, which analyzed different sounds, such as sound bombs and air attacks.
After deducting an approximate position for the place where Mr. Biari was making his calls, Israeli military officials were warned that the area, which included several apartments of apartments, was densely populated, said two intelligence officers. An aerial attack should target several buildings to ensure that Mr. Biari was assassinated, they said. The operation was illuminated in green.
Since then, Israeli intelligence has also used the audio tool together with maps and photos of the Gaza underground tunnel labyrinth to identify the hostages. Over time, the tool has been perfected to find more precisely individuals, two Israeli officers said.