.
AI models are already deployed in societies affected by armed conflict, and journalists, humanitarian workers, governments and ordinary citizens rely on them for information or for their work processes. No established practice exists for checking whether their outputs can make those conflicts worse. We tested nine model configurations from four providers (OpenAI, Anthropic, DeepSeek, xAI) on 90 multi-turn scenarios designed to surface misaligned behaviour in conflict contexts: false equivalence between documented atrocities, denial of genocide, and failure to recognise ethnic slurs, among others. When such outputs feed into journalism, humanitarian reporting, or public debate, they can deepen divisions in fragile societies. Failure rates span 6% to 47% between the best and worst performing models, which makes model choice a safety question in its own right and when users pushed for “balance” in cases where international courts have already assigned responsibility, five of nine configurations failed 80 to 100 percent of the time. We release the first evaluation framework for this domain and propose adding it to alignment evaluation portfolios.
The aim was to compare clinical and radiographic outcomes between myelomeningocele (MMC) and other neuromuscular early-onset scoliosis (EOS) patients following graduation from growing rod treatment and to evaluate whether MMC should be considered a distinct subgroup within the classification of EOS. This single-center retrospective cohort study included 26 patients with neuromuscular EOS (10 with MMC, 16 with other neuromuscular etiologies) treated with traditional or magnetically controlled growing rods between 2004 and 2023, all of whom achieved graduation and had a minimum of 2-year of post-graduation follow-up. Demographic, clinical, radiographic, and complication data were analyzed. Complication burden was assessed using unplanned return to the operating room (UPROR), modified Clavien–Dindo–Sink grading, and negative binomial regression. Wound-related complications were significantly more frequent in the MMC group (80.0
Receptive labeling, also referred to as auditory-visual conditional discrimination (AVCD), is commonplace in everyday interactions; yet, many individuals with autism spectrum disorder experience barriers to learning AVCD during instruction. When clients fail to make progress during AVCD instruction, behavior analysts may engage in a trial-and-error process for intervention modifications that may be inefficient or unsuccessful. This paper presents a problem-solving model to identify and address learning barriers during AVCD instruction. The model outlines eight steps encompassing data collection, instructional and procedural analysis, prerequisite skill evaluation, hypothesis development, barrier assessment with matched interventions, intervention selection, and progress monitoring. This approach provides behavior analysts with practical guidance for clinical decision-making, moving toward systematic, assessment-based intervention that can improve outcomes for clients receiving AVCD instruction.
We consider a game theoretic approach to solve multi-agent coordination problems with submodular maximization objectives. It is known for such problems that the Nash equilibria for the corresponding game are always within 50
Existing AI agent safety benchmarks focus on generic criminal harm (cybercrime, harassment, weapon synthesis), leaving a systematic blind spot for a distinct and commercially consequential threat category: agents harming their own deployers. Real-world incidents illustrate the gap: Slack AI credential exfiltration (Aug 2024), Microsoft 365 Copilot calendar-injection leaks (Jan 2024), and a Meta agent unauthorized forum post exposing operational data (Mar 2026). We propose Owner-Harm, a formal threat model with eight categories of agent behavior damaging the deployer. We quantify the defense gap on two benchmarks: a compositional safety system achieves 100
