Policy
Laws, arrests, takedowns
India proposes DEPA framework to train AI models on local data without moving it
India's Ministry of Electronics and Information Technology has outlined the DEPA for AI Training framework, in which Training Data Providers such as hospitals and banks share anonymised, encrypted data with Data Exchange Platforms. Training Data Consumers then train models inside hardware-based secure data environments, with differential privacy and immutable audit logging. The framework is a proposal and its implementation timeline is unclear.
Court cases over US 2026 midterm election integrity and federal voter data actions
Just Security's tracker lists active 2026 midterm cases, including US government suits against California and Alaska and challenges to DHS and DOJ voter-list and enforcement actions. Recent rulings include a preliminary injunction barring DHS election-related field operations in nine states, a temporary restraining order limiting DOJ and DHS threats against election officials, and the vacating of a federal Voter List Maintenance Policy. Several hearings and decisions are scheduled for early to mid November 2026.
Just Security collection covers legal and policy issues around the 2026 US midterm elections
Just Security has published a topic-organised collection of 2026 US midterm analysis covering election administration, federal seizures of election records, troops at polling places, USPS mail ballot litigation, redistricting and voting rights, political violence and AI-driven foreign influence. It also points readers to a continually updated election integrity litigation tracker for legal challenges to administration actions.
Amazon memo on AI use in engineering draws scepticism over mixed messages and token-usage metrics
An Amazon memo by a Distinguished Engineer says engineers who find AI does not help their work should document it as a finding, not a failing. The columnist argues the memo contradicts itself on input metrics and warns that token-usage targets, like Amazon's shut-down KiroRank leaderboard, can be gamed, while AWS earns revenue from token billing.
EFF Warns Narrowing Fair Use for Generative AI Training Would Harm the Public
The EFF has filed an amicus brief in In re OpenAI, arguing that narrowing the fair use doctrine would affect far more than generative AI developers. It warns that requiring lawful acquisition of training data or mandatory licensing would entrench large tech firms, impede research and journalism, and cast doubt on routine copying in everyday internet use.
Judge finds USS Cole bombing defendant competent to stand trial at Guantanamo
An Army judge at the Guantanamo Bay military commission ruled that Abd al-Rahim al-Nashiri is competent to stand trial, rejecting a defence claim that torture and his conditions had left him unfit. Jury selection is scheduled to start on 2 November and could take weeks, with about 200 service members expected to be brought to the base in groups. The case is the first capital case before the commission.
Identity governance becomes an executive accountability issue as AI agents and machine identities multiply
The article says human and nonhuman identities, including service accounts, machine identities and AI agents, are changing faster than manual access reviews can track. It cites Foundry research ranking identity security among the top cybersecurity priorities for 2026 and notes that security leaders may face personal accountability after a breach or compliance failure. It recommends continuous identity security: unified discovery of identities, always-on lifecycle management, policy enforcement and access validation, and retained evidence of governance activity for auditors.
Greg Touhill discusses using modelling and simulation to aid cyber operators
Federal News Network's cybersecurity section carries a piece titled "Every cyber threat can't be a top priority". The only quoted material is Greg Touhill saying modelling and simulation can help cyber operators cope with the "Fog of Cyberwar". The excerpt provided gives no further detail on methods or outcomes.
GM agrees $220 million settlement with 36 attorneys general over driver data sold to brokers
GM has settled with 36 attorneys general for $220 million over collecting driving data, including precise location, speed and seatbelt use, via its OnStar Smart Driver programme and passing it to brokers that supplied insurers. Investigators said dealership staff sometimes accepted terms on customers' behalf, and the OnStar app wrongly told more than 370,000 drivers they were not enrolled. GM also agreed to limit the data it collects, better protect customer PII and set up a comprehensive data privacy programme, following an FTC order finalised in January 2026 and a separate $12.75 million California settlement in May.
FBI involved in response to $6.3 billion Binance hack linked to North Korea
The FBI is involved in the response to a hack against the cryptocurrency exchange Binance, reported as $6.3 billion and linked to North Korea. The source headline gives no further detail on the method, timing or recovery of funds.
New RAID-AI framework focuses on ransomware resilience for AI systems
Infosecurity Magazine reports on RAID-AI, a new framework intended to help organisations assess their security preparation for ransomware attacks that target AI systems. The article excerpt gives no further detail on its structure, authors or requirements.