Large vision-language models (VLMs) are highly vulnerable to jailbreak attacks that exploit visual-textual interactions to bypass safety guardrails. Rather than relying on curated safety-specific data or costly image-to-text conversion, we introduce a new formulation of the safety-relevant distributional shift induced by the visual modality. Large vision-language models (VLMs) (e.g., LLaVA Liu et al. (2023), InternVL Chen et al. (2024), and MiniGPT Zhu et al. (2024)) integrate vision and language capabilities, achieving remarkable multimodal modeling performance. However, incorporating visual modality introduces new vulnerabilities, making VLMs more susceptible to malicious manipulations than their backbone language models Liu et al. (2024). In multimodal jailbreaks, adversaries exploit the intricate interactions between visual and textual inputs to circumvent target VLMs' safety guardrails and elicit harmful responses Qi et al. (2023). A variety of attacks have been proposed, such as pairing harmful text with adversarially perturbed images Li et al. (2024), and embedding harmful content into images via generative models Liu et al. (2024) or typography Jiang et al. (2025).

The remarkable success of Large Language Models (LLMs) has illuminated a promising pathway toward achieving Artificial General Intelligence for both academic and industrial communities, owing to their unprecedented performance across various applications. As LLMs continue to gain prominence in both research and commercial domains, their security and safety implications have become a growing concern, not only for researchers and corporations but also for every nation. Currently, existing surveys on LLM safety primarily focus on specific stages of the LLM lifecycle, e.g., deployment phase or fine-tuning phase, lacking a comprehensive understanding of the entire "lifechain" of LLMs. To address this gap, this paper introduces, for the first time, the concept of "full-stack" safety to systematically consider safety issues throughout the entire process of LLM training, deployment, and eventual commercialization. Compared to the off-the-shelf LLM safety surveys, our work demonstrates several distinctive advantages: (I) Comprehensive Perspective. We define the complete LLM lifecycle as encompassing data preparation, pre-training, post-training, deployment and final commercialization. To our knowledge, this represents the first safety survey to encompass the entire lifecycle of LLMs. (II) Extensive Literature Support. Our research is grounded in an exhaustive review of over 800+ papers, ensuring comprehensive coverage and systematic organization of security issues within a more holistic understanding. (III) Unique Insights. Through systematic literature analysis, we have developed reliable roadmaps and perspectives for each chapter. Our work identifies promising research directions, including safety in data generation, alignment techniques, model editing, and LLM-based agent systems. These insights provide valuable guidance for researchers pursuing future work in this field.

Olive production is an important tree crop in Mediterranean climates. However, olive yield varies significantly due to climate change. Accurately estimating yield using remote sensing and machine learning remains a complex challenge. In this study, we developed a streamlined pipeline for olive yield estimation in the Kairouan and Sousse governorates of Tunisia. We extracted features from multispectral reflectance bands, vegetation indices derived from Landsat-8 OLI and Landsat-9 OLI-2 satellite imagery, along with digital elevation model data. These spatial features were combined with ground-based field survey data to form a structured tabular dataset. We then developed an automated ensemble learning framework, implemented using AutoGluon to train and evaluate multiple machine learning models, select optimal combinations through stacking, and generate robust yield predictions using five-fold cross-validation. The results demonstrate strong predictive performance from both sensors, with Landsat-8 OLI achieving R2 = 0.8635 and RMSE = 1.17 tons per ha, and Landsat-9 OLI-2 achieving R2 = 0.8378 and RMSE = 1.32 tons per ha. This study highlights a scalable, cost-effective, and accurate method for olive yield estimation, with potential applicability across diverse agricultural regions globally.

Amongst the most common use cases of modern AI is LLM chat with web search enabled. However, no direct evaluations of the quality of web research agents exist that control for the continually-changing web. We introduce Deep Research Bench, consisting of 89 multi-step web research task instances of varying difficulty across 8 diverse task categories, with the answers carefully worked out by skilled humans. We provide a "RetroSearch" environment with a large frozen set of scraped web pages, and demonstrate that offline "RetroSearch" agents perform comparably to "live web" agents, enabling reliable evaluations of models over time. We provide robust agent tooling and scaffolding to benchmark major LLMs as they are released, including "thinking" models like o3 and Gemini 2.5 Pro. We include automated evaluations of the lengthy agent traces to report progress over time in hallucinations, tool use, and forgetting. Finally, we evaluate the major web research products branded as "Deep Research", "Deep Search", "Search", or "Research." Results are available on a public leaderboard at https://drb.futuresearch.ai/.

Thousands of demonstrators took to the streets of Los Angeles County over the weekend to protest immigration raids by Immigration and Customs Enforcement. In response, federal, state, and local authorities have all escalated a brutal crackdown against the anti-ICE demonstrators, with law enforcement deploying tear gas, bean bags, and rubber bullets against protesters and journalists alike, and President Donald Trump activating state National Guard members in defiance of Governor Gavin Newsom's wishes. "This is a chilling statement." On Sunday, according to the Los Angeles Times, the Los Angeles Police Department made another escalation, flying over a group of demonstrators in downtown LA in a helicopter and announcing, "I have all of you on camera. I'm going to come to your house."

Craig Federighi's opening remarks were nearly derailed by a protester who snuck into WWDC 2025 Monday afternoon. It's not clear what the protester said while they tried to interrupt Federighi's presentation of iOS 26, CNBC reported. Several major technology events have been interrupted by protests recently, most of which supported Hamas and protested the tech industry's business ties to the Israeli government. Apple CEO Tim Cook closed out Monday presentation of the company's new operating system by announcing that iOS 26 is now available to developers for Beta testing today. That means a pre-release version of the iOS 26 operating system for iPhones is now accessible for registered Apple developers to test.

Fox News' Jonathan Hunt reports the latest on the anti-ICE riots in Los Angeles. Correspondent Rich Edson details Dems' response to Trump deploying the National Guard and'Outnumbered' co-host Kayleigh McEnany weighs in on the escalation. Dozens of protesters have been arrested following a weekend of violence across Los Angeles as tensions hit a boiling point over immigration raids throughout the city. On Sunday, law enforcement officials from multiple agencies arrested 41 protesters as anti-Immigration and Customs Enforcement (ICE) demonstrations spiraled out of control. Of the nearly four-dozen arrests, 21 were made by the Los Angeles Police Department (LAPD), 19 by California Highway Patrol and one by the Los Angeles Sheriff's Department.

Top US and Chinese officials are meeting in London in a bid to defuse trade tensions over rare earth minerals and advanced technology after a phone call between Presidents Donald Trump and Xi Jinping last week. The two sides are aiming in Monday's talks to build on a preliminary trade deal struck in Geneva in May, which briefly lowered the temperature between Washington and Beijing and offered relief for investors battered by months of Trump's global trade war. Since then, the agreement to mutually suspend most of the 100 percent-plus tariffs for 90 days has been followed by barbs and accusations from both sides. But after reaching a tentative understanding with Xi on resuming the flow of critical minerals, Trump said on Thursday that he expected Monday's meeting to go "very well". The US delegation in London is headed by Treasury Secretary Scott Bessent, Commerce Secretary Howard Lutnick and Trade Representative Jamieson Greer.

Russia has launched 479 drones against Ukraine in the biggest overnight drone bombardment of the three-year war, according to the Ukrainian air force. The air force said early on Monday that it had downed 460 drones as well as 19 missiles launched overnight. Russia's continued to step up its drone and missile attacks on Ukraine, despite declaring, under pressure from United States President Donald Trump, that it is interested in pursuing peace talks. The record launch came just ahead of the start of a prisoner swap agreed at recent talks between the pair. Of the hundreds of projectiles fired at numerous targets, only 10 reached their destination, Kyiv officials said.

All civil servants in England and Wales will get practical training in how to use artificial intelligence to speed up their work from this autumn, the Guardian has learned. More than 400,000 civil servants will be informed of the training on Monday afternoon, which is part of a drive by the chancellor of the duchy of Lancaster, Pat McFadden, to overhaul the civil service and improve its productivity. At the same time, the size of the civil service is being reduced by tens of thousands of roles through voluntary redundancy and not replacing leavers. The government said officials would be tasked with figuring how they could use AI technology to streamline their own work wherever possible. Officials are already piloting a package of AI tools called Humphrey – named after the senior civil servant Sir Humphrey Appleby from the 1980s TV sitcom Yes, Minister.