Amnesty: AI surveillance dangers ‘supercharging’ US deportations
Automated synthetic intelligence (AI)-powered surveillance instruments are being deployed to trace migrants, refugees and asylum seekers within the US, elevating critical human rights considerations, in response to a report by Amnesty Worldwide.
Amnesty’s evaluation of paperwork obtained from the Division of Homeland Safety (DHS) highlights how two methods specifically – Babel X, supplied by Babel Road, and Palantir’s Immigration OS – have automated monitoring and mass surveillance capabilities which can be getting used to underpin the authorities’s aggressive immigration enforcement operations.
The organisation claims the instruments feed into the State Division’s AI-driven “Catch and Revoke” initiative, which mixes social media monitoring, visa standing monitoring and automatic menace assessments of overseas people on visas. The observe has already been criticised for violating the First Modification rights of individuals residing within the US.
Amnesty warns that the velocity and scale at which these applied sciences can determine individuals and infer their behaviour may lead to mass visa revocations and deportations.
“It’s deeply regarding that the US authorities is deploying invasive AI-powered applied sciences inside a context of a mass deportation agenda,” mentioned Erika Guevara-Rosas, senior director for analysis, advocacy, coverage and campaigns at Amnesty Worldwide.
“The coercive Catch and Revoke initiative, facilitated by AI applied sciences, dangers supercharging arbitrary and illegal visa revocations, detentions, deportations and violations of a slew of human rights.”
The instruments
Babel X, a data-mining platform developed by Babel Road, has been utilized by US Customs and Border Safety (CBP) since not less than 2019. It collects huge quantities of private info, together with names, emails, telephone numbers, IP addresses, employment data and cellular promoting IDs that reveal gadget places. The instrument also can monitor social media posts.
Amnesty says this info is fed into AI methods that scan social media for “terrorism”-related content material, which might then be used to resolve whether or not a person’s visa needs to be revoked. As soon as a visa is revoked, Immigration and Customs Enforcement (ICE) brokers could be dispatched to deport the particular person in query.
Palantir’s Immigration Lifecycle Working System (Immigration OS) was launched following a $30m contract with ICE in April 2025. The system integrates datasets throughout companies, enabling ICE to construct digital case recordsdata, hyperlink investigations and observe private info on immigrants. Its up to date options embrace streamlining arrests based mostly on ICE priorities, real-time monitoring of “self-deportations” and figuring out precedence deportation instances, significantly visa overstayers.
In line with Amnesty, the usage of such instruments has been crucial in enabling US authorities to scale up deportations. Nonetheless, the organisation warns in addition they improve the chance of illegal actions by drawing on a number of private and non-private sources with out enough oversight.
The non-governmental organisation contacted each corporations, with Babel not offering any remark, and Palantir stating that its product was not used to energy the administration’s Catch and Revoke effort.
The report additional notes that probabilistic methods similar to these usually depend on behavioural inferences, which could be discriminatory. For instance, pro-Palestine content material might be falsely categorised as antisemitic, amplifying current biases.
“Algorithms are socially constructed, and our world is constructed on systemic racism and historic discrimination,” mentioned Petra Molnar, a lawyer specialising in migration and human rights and director of the Refugee Legislation Lab at York College. “These instruments are going to duplicate biases already inherent to the immigration system, to not point out create new ones based mostly on very problematic assumptions about human behaviour.”
Molnar careworn that there’s an underlying layer of “systemic discrimination that undercuts all of this” based mostly on the belief that “individuals on the transfer are in some way a menace”.
“That is in the end about dehumanisation. That’s the central narrative that the Trump administration is pushing,” she mentioned. “There was an exponential improve in the kind of applied sciences and surveillance mechanisms which can be being more and more weaponised in the direction of individuals on the transfer and cellular communities.”
Amnesty additionally criticises Palantir and Babel Road for failing to hold out enough human rights due diligence, arguing that corporations are liable for guaranteeing their applied sciences should not deployed in ways in which violate human rights.
Molnar pointed to the Ruggie Ideas, a UN framework that units out company duties on this space: “That is an impartial normal for personal corporations. They’ve to stick to worldwide authorized rules with regards to the event and deployment of expertise.”
For Molnar, the best answer would contain a “strong human-rights respecting framework”, together with human rights and information affect assessments performed all through the lifecycle of a challenge. However she careworn the necessity for “public consciousness of what these corporations are doing” and “a divestment from sure corporations”.
“There must be an open dialogue between individuals who truly develop the expertise and the affected neighborhood, as a result of there’s this wall proper now between individuals who develop the tech and the individuals who the tech is hurting,” she mentioned.
“These are traits I’ve been seeing internationally. It’s not simply in the USA, however I feel the USA is the newest manifestation.”
Laptop Weekly contacted Palantir and Babel Road in regards to the considerations raised by Amnesty’s report, and requested quite a few additional questions, together with how the companies are working to cut back algorithmic bias, the measures they’re taking to keep away from detrimental human rights impacts with their deployments, and whether or not both firm has consulted with affected migrant communities.
Neither had responded by the point of publication.
UK parallels
Comparable patterns are rising within the UK. Human rights campaigners on the Migrants’ Rights Community (MRN) have investigated the usage of AI on the border and highlighted its rising position in surveillance applied sciences similar to facial recognition.
“AI applied sciences are used underneath the guise of effectivity. It permits border immigration methods to turn out to be automated. It reduces the necessity for human intervention, for borders to be reliant on patrols or bodily partitions,” mentioned an MRN consultant.
The organisation argues that authorities reliance on non-public contractors dangers creating and aggravating an already “digital hostile setting”. However, they add, it’s usually tough to acquire details about how these applied sciences are getting used.
For instance, when investigating the Residence Workplace’s deployment of Anduril Maritime Sentry Towers on the south-east coast of England, researcher Samuel Storey needed to file 27 separate Freedom of Info (FOI) requests. Whereas the Residence Workplace claimed the towers have been supposed to help “environmental safety”, Storey argues they’re getting used for surveillance of migrant crossings between the UK and France.
“The FOI system is an extension of state secrecy. It’s probably not a instrument for the liberty of data, however an extension of the state’s capability to not expose or disclose,” he mentioned.
MRN has additionally raised questions on information entry and privateness.
“It’s been extremely tough to search out out, however that information can be saved someplace, and we’ve a suspicion that it will likely be saved in one among these Amazon Net Companies hubs, as a result of they’ve large contracts with the federal government. Private information is likely one of the most precious issues that corporations can have,” mentioned the consultant.
The priority, they added, is that if non-public corporations retailer the information, these corporations would have entry to this information, and extra entities, similar to overseas governments, could technically have the ability to entry it too.
Earlier reporting by Laptop Weekly warned that the brand new eVisa methods might be used to trace migrants and help immigrant enforcement.
