Licence to abuse? The problem with Israel’s signing of global treaty on AI

On September 5, Israel became one of 57 countries to sign the world’s first global treaty on artificial intelligence, which purports to establish a legal framework to harness AI for the “common good” while safeguarding “democracy and human rights”.

While Israel's participation in this treaty may seem like a step toward greater accountability, it stands in stark contrast to its ongoing deployment of AI systems that have been widely criticised for infringing on the basic human rights of Palestinians.

The AI treaty, spearheaded by the Council of Europe, mainly focuses on the protection of human rights of people affected by AI systems, yet, its provisions concerning national security and the private sector remain vague, which raises significant concerns about the potential for Israel to circumvent the treaty's principles.

The treaty is “a fairly loose statement of principles, although it might be legally binding,” Andreas Krieg, a senior lecturer at the School of Security Studies at King's College London, tells TRT World.

“Quite a lot of variation” is to be expected with regard to how different states could interpret and apply these internationally agreed principles into applicable laws, he notes.

The treaty’s text states that the Framework Convention “does not apply to national defence matters nor to research and development activities,” except when AI testing might interfere with human rights, democracy, or the rule of law.

This creates a grey area where it is unclear when and how the treaty’s principles should be enforced, particularly in military-related AI developments, according to experts.

AI plays a significant role in the kill chain in Israel’s war on Gaza, which potentially violates the convention, and for this reason alone, Kreig feels that Tel Aviv will not invoke the treaty when it comes to “any sort of military operation in the name of national security”.

Israel’s war on Gaza has by now killed more than 41,000 people, mostly women and children, which Krieg underscores is “a direct result of AI targeting systems that are not transparent”.

In April, the Lavender, an AI-based target creation tool used by the Israeli military, made headlines for facilitating civilian casualties of catastrophic proportions in the Palestinian enclave.

According to a report by +972 Magazine, the system’s algorithm indicated the acceptance of 15-20 civilian casualties for one low-ranking Hamas member and up to 100 for one senior Hamas member, making recommendations based on behavioural “features”, including communications patterns, social media connections, and changing addresses frequently.

How data is collected by AI systems such as the Lavender remains unclear, but experts suggest that sources could include drone footage, intercepted private communications as well as surveillance data, which raises serious concerns about transparency.

Al Halil, a city in the southern part of the occupied West Bank, is one of the many areas where similar technologies have been employed.

Across the occupied city, Palestinian residents’ faces are constantly scanned by CCTV cameras paired with AI-driven software that collects vast amounts of personal data on them, raising significant concerns about human rights, as detailed in an Amnesty International report published in 2023.

“We don’t know what they collect exactly,” says Issa Amro, a Palestinian human rights defender from Al Halil.

“Maybe they use us as a simulation for their technologies and for researchers. Maybe they sell our data to labs around the world without our consent,” Amro tells TRT World.

Israel justifies extensive surveillance and movement restrictions in the occupied West Bank and East Jerusalem under the pretext of national security, raising significant questions about the meaningfulness of the treaty’s provisions of national security exemptions.

“The liberal principles of Israeli law usually apply primarily or exclusively to Israeli citizens and not to Palestinians,” Krieg says.

Israel will likely continue using AI in the name of national security, Krieg maintains. “So none of these principles to do with transparency or respect for human rights will be applied to occupation.”

The treaty also does not lay out specific, enforceable guidelines or mechanisms for regulating the private sector's use of AI.

Instead, it leaves the responsibility of addressing risks arising from AI activities by private actors to individual nations, which could lead to significant variations in how strictly the private sector is regulated.

Krieg notes that this could result in governments increasingly delegating the development and weaponisation of AI to the private sector to bypass the principles of the treaty and avoid being held to account.

He warns that this could lead to an increased lack of transparency, as it would leave space for private contractors to conduct operations with AI without being forthcoming about how these systems function, effectively shifting the responsibility of compliance from governments to private entities.

Many of Israel's destructive AI systems are typically developed by private defence contractors or technology companies in collaboration with the Israeli government or military.

For instance, in late 2022, the Israeli army confirmed that it was testing a remote-controlled riot control tool that could fire shots, developed by a private firm Smart Shooter.

Here is a closer examination of some of the controversial AI softwares Israel has employed in its surveillance and military actions against Palestinians:

Mabat 2000

Mabat 2000 is an extensive facial recognition system deployed by Israeli authorities in occupied East Jerusalem.

Used in tandem with CCTV cameras, this technology is designed to identify protesters and keep Palestinians under constant surveillance, even as they go about their daily lives.

In a 2023 report by Amnesty International, Neda, a Palestinian resident of occupied East Jerusalem, described how this invasive technology impacts her daily life:

The Lavender

The Lavender is a data processing system developed following October 7, which resulted in a list of 37,000 bombing targets, according to an investigation by Israel-based publications +972 Magazine and Local Call.

The algorithm reportedly indicated the acceptance of 15-20 civilian casualties for one low-ranking Hamas member and up to 100 for one senior Hamas member, making targeting recommendations based on behavioural “features”, including communications patterns, social media connections, and changing addresses frequently.

Wolf Pack

Wolf Pack is a comprehensive database that compiles extensive information on Palestinians from the Occupied Palestinian Territories (OPT), including details such as their residences, family members, and whether they are wanted for questioning by Israeli authorities, according to Amnesty International.

The database operates in conjunction with facial recognition systems.

Red Wolf

"Red Wolf" is a facial recognition system that has been deployed at Israeli checkpoints in Al Khalil since at least 2022, according to Amnesty International.

When Palestinians pass through these checkpoints, their faces are scanned without their knowledge or consent and matched against a database containing only Palestinian profiles.

Red Wolf decides if they can pass or be denied entry and automatically enrols new faces it scans.

The system can also block passage based on stored information, such as whether an individual is wanted for questioning or arrest.

Blue Wolf

Introduced in 2020, Blue Wolf is a gamified app often referred to as the "Facebook of Palestinians." It allows Israeli soldiers to upload and view photos, locations, and detailed profiles of Palestinians entering the city.

Accessible via smartphones and tablets, Blue Wolf can instantly retrieve information from the extensive Wolf Pack database, giving Israeli forces real-time access to data on individuals.

The Gospel

The Gospel is an AI-driven target-creation platform described by Israeli operatives as a “mass assassination factory.”

In an interview published before October 7 in the Israeli newspaper Ynet, Aviv Kochavi, the former head of the Israeli military, revealed that during Israel’s 11-day war on Gaza in May 2021, once the Gospel was activated, it generated 100 targets a day.

"To put that into perspective, in the past, we would produce 50 targets in Gaza per year. Now, this machine produces 100 targets in a single day, with 50% of them being attacked."

Persistent lack of red lines for mass killings

Concerns about ambiguity and the absence of clear “red lines” in the world’s first international AI treaty have persisted across previous drafts of the Council of Europe Framework Convention on AI, drawing significant criticism from civil society groups and academics.

In an open letter addressed to the European AI convention negotiators in March, dozens of civil society organisations argued that the treaty provides a “free pass” to tech and security companies.

They called on the State Parties and the EU to “not water down our rights,” urging the rejection of blanket exemptions related to national security and defence.

“Some states are also pushing for a blanket exemption with regard to national security and defence,” the letter read: “Nothing justifies the unconditional waiving of the safeguards set in international, European and national law that usually apply in these fields.”

One of the signatories of the letter was Digital Action, an international advocacy group established to protect democracy and human rights from digital threats.

Mona Shtaya, an advocacy and digital communications expert from the group, expresses that even before October 7, Israel had been systematically working to automate its apartheid systems in the occupied Palestinian territories.

“It’s profoundly ironic that Israel, which weaponises AI to massively and indiscriminately target and kill Palestinians, now celebrates joining the global AI convention,” Mona Shtaya tells TRT World.

“From the so-called smart shooter in Al Halil to facial recognition technologies, it has weaponised groundbreaking technologies to target and kill Palestinians. Yet, it is still allowed to sign ‘the first global convention on artificial intelligence.’”