News Briefing 12-16 Feb 2024

Autonomy in weapons systems, and AI in security and defence

• US: DIU consults Pentagon policy shop to move Replicator technology at ‘responsible speed’: The Defense Innovation Unit, responsible for the U.S. Department of Defense’s Replicator programme (which aims to create ‘thousands of weapons systems powered by artificial intelligence’, as covered in numerous news briefings last year), are reportedly ‘engaging deeply with the Pentagon’s policy team regarding the application of DOD Directive 3000.09, which lays out the rules and process for approving the acquisition and deployment of lethal autonomous platforms that can engage military targets without troops intervening.’ Aditi Kumar, the Defense Innovation Unit’s deputy director of strategy, policy and national security partnerships, said that ‘we want to be on the cutting-edge of what’s possible, but we want to do that in line with our values and our tenets for responsible use. And the policy, what it says is not that there needs to be a human in the loop, but that there needs to be human judgment in the deployment of these weapon systems.’
• Russia: Russia’s Automated Killer Drones May Not Be Working As Planned: Writing in Forbes, David Hambling reports that Russia’s Lancet drone, which had been reported to have certain automated functionalities, ‘has not worked as well as planned, and the automated mode may have been disabled’. He further writes that ‘the obvious conclusion is that automated target recognition software was rolled out prematurely and there has been product recall.’
• US: Northrop to showcase Project Scion tech payloads for unmanned vessels: Defense company Northrop Grumman will ‘participate in two events to demonstrate autonomy and electronic warfare payloads the company is developing for unmanned surface vessels under its Project Scion initiative.’
• US: Navy robot ships on a 15-year path to operating ‘at speed and scale,’ CNO says: DefenseOne carries details and quotes from U.S. Navy and industry figures on the Naval Service’s pursuit of autonomous and unmanned systems.

Regulation and enforcement

• EU: Artificial Intelligence Act: committees confirm landmark agreement: A press release issued by the European Parliament this week revealed that ‘MEPs at a committee level have endorsed the provisional agreement on the Artificial Intelligence Act –  that ensures safety and complies with fundamental rights.’ The agreement will be formally adopted in an ‘upcoming Parliament plenary session and It will be enforced within 24 months after entry-into-force.’

Facial recognition, biometric identification, surveillance 

• UK: London Underground Is Testing Real-Time AI Surveillance Tools to Spot Crime: Matt Burgess, writing for the WIRED, reports that a test was recently conducted at one of London underground stations to try and detect acts of crime, weapons use and people possibly falling off the railway tracks.
• Ireland: Ireland to hear from stakeholders on police facial recognition bill: This week a number of stakeholders convened to examine ‘the draft Garda Síochána (Recording Devices) (Amendment) bill, which was approved by the Irish Cabinet at the end of 2023.’ The author notes that ‘the Data Protection Commission plans to say that while facial recognition may facilitate policing, “this technology presents serious risks to the individual’s right to data protection,” and that such technology should only be used when it “is strictly necessary and proportionate.” This comes at a time where there have been cases worldwide in which individuals have been wrongfully identified by facial recognition technology, and arrested.
• New Zealand: Facial recognition raises dilemma for police: Waatea News reports on comments by Deputy Police Commissioner Tania Kura which highlight the struggles faced by the department in using facial recognition technologies. She suggests an ‘open discussion’ on this issue and calls for a balance between ‘efficiency’ and individual rights/ freedoms. 
• International: Artificial Intelligence Is Putting Innocent People at Risk of Being Incarcerated: Alyxaundria Sanford writes for the Innocence Project on documented wrongful conviction of black people due to misidentification by facial recognition systems. The piece questions the claims of accuracy made by the manufacturers of these systems, and calls for reliability and accountability. 
• UK: Met Police nab 45 people with facial recognition, deployments coming to Lincolnshire: Biometric Update reports on the deployment of live and retrospective facial recognition technologies in London by the Metropolitan Police which led to the arrest of approximately 45 people since it began operations. Chief Superintendent Andy Brittain has hailed the accuracy of the system which is 1 in 6000 false positive rate, despite academics and experts calling out the lack of ‘clear legislation or democratic debate’ in deploying such systems.

AI, algorithms and autonomy in the wider world

• US: Bring on the AI guardrails!: This piece in Berkley Public Health covers comments by Ziad Obermeyer, Blue Cross Distinguished Associate Professor of Health Policy and Management at Berkeley Public Health during a senate hearing titled ‘Artificial Intelligence and Health Care: Promise and Pitfalls.’ He notes that while AI prediction tools have been helpful for medical professionals, there is risk of ‘automating and scaling up racial bias.’ 
• US: FCC bans AI-generated voices in robocalls that can deceive voters: In this piece by PBS NewsHour, it reports on The Federal Communications Commission’s (FCC) ban on ‘robocalls that contain voices generated by artificial intelligence.’ This ruling by the FCC declares AI generated calls as artificial’ under the Telephone Consumer Protection Act, a 1991 law ‘restricting junk calls that use artificial and prerecorded voice messages.’ 
• International: AI will not replace humans any time soon: Meta scientist: ET CIO reports that Dr Yann LeCun, Vice President and Chief AI Scientist at Meta debunked exaggeration of AI abilities during a plenary session at the World Governments Summit (WGS) 2024. He notes that the present cognitive abilities of AI are ‘less sophisticated than those of cats and dogs’ and lack ‘planning capabilities’ as well as ‘speed or temporal accuracy to analyse and manage complex issues.’