Results for “Prompt”

The Guardian World Lauren Almeida and Jasper Jolly

BAE to slash 1,000 jobs across head office and air business to streamline orders

Job cuts to exclude shop floor, aircraft maintenance and trainees but move prompts calls for more orders to support British defence giant BAE Systems has announced plans to cut more than 1,000 jobs from its workforce, triggering calls for ministers to place extra orders with Britain’s biggest defence contractor. The company said it is aiming to cut 1,070 jobs at its head office and air business in the UK, as it tries to meet customer demands for “greater speed and efficiency…

Read at The Guardian World

arXiv (research papers) Almene De Meran Meguimtsop SI

Why Do LLMs Help with Scientific Misconduct Requests and What to Do About It?: Increasing the Trustworthiness of AI for Science

Large language models (LLMs) support scientific research, but some of that assistance can also undermine research integrity. Existing safety evaluations in science often measure whether models refuse misconduct requests, but rarely explain whether they even recognize the problem or why they comply. We investigate this gap across 8 LLMs using SciIntBench, an existing benchmark of 1,350 prompts about scientific research integrity. We examine the reasons behind compliance in…

Read at arXiv (research papers)

arXiv (research papers) BoHao Chen SI

Beyond Weight Updates: Evolving Agent Harnesses for Domain Specialization

Domain specialization of LLM agents is usually framed as a weight-update problem: collect trajectories, run reinforcement learning (RL) or supervised fine-tuning (SFT), and deploy a new checkpoint. We study a complementary axis: evolving the \textbf{agent harness}---the system prompt, tool schemas, skill library, and planning scaffolding---while keeping the policy frozen. We cast harness evolution as an RL-style outer loop and instantiate it with GEPA, a reflective…

Read at arXiv (research papers)

MacRumors Juli Clover Gadgets

Instagram Now Lets You Save Songs to Apple Music

Sometime in the last week or two, Instagram added a new Apple Music feature. You can now connect an ‌Apple Music‌ account to the social network and save songs you come across to an ‌Apple Music‌ playlist. If you tap a song link on a post or a story on Instagram, you'll see a prompt to connect Instagram to ‌Apple Music‌. Once your ‌Apple Music‌ account is linked to Instagram, you can tap a song name in a post and use the "Add to…

Read at MacRumors

The Hill Julia Shapero Geopolitics

How AI actually played a role in recent tech outages

Presented by Google Public Sector. {beacon} View Online Technology Technology The Big Story How AI actually played a role in recent tech outages A wave of technology outages and issues across the US prompted the public to quickly point fingers at AI, but it might've not been used in the way people thought. © AP...

Read at The Hill

PBS News Politics Josh Funk, Associated Press Geopolitics

DOT Secretary Sean Duffy keeps blaming L3Harris after FAA ground stop, even as cause remains unknown

U.S. Transportation Secretary Sean Duffy continues to blame a contractor for the events that prompted the Federal Aviation Administration to order a ground stop last week, even though investigators haven't determined the root cause.

Read at PBS News Politics

arXiv (research papers) Syed Ahzaz Shah Security

PriTabICL: Differentially Private Tabular In-Context Learning with Task-Aligned Cluster-Wise Synthesis

Tabular in-context learning (TabICL) enables large language models (LLMs) to predict target records using a handful of labeled demonstrations in their prompts. However, these demonstrations often contain sensitive personal information. Differential privacy (DP) offers principled protection, but noise-based methods introduce substantial noise, distorting the feature-label relationships essential for accurate prediction, especially as data dimensionality grows. In this work…

Read at arXiv (research papers)

WIRED AI Matt Burgess SI

Criminals Have a New Favorite ‘Point-and-Shoot’ AI Hacking Tool

Artex and similar penetration testing tools allow newbie hackers to scan targeted systems for vulnerabilities and successfully wage attacks using prompts fed to chatbots.

Read at WIRED AI

The Decoder Maximilian Schreiner SI

Anthropic, OpenAI, and Google models readily marked medics, journalists, and surrendering soldiers as kamikaze drone targets

A Vals AI investigation shows that leading AI models from Anthropic, OpenAI and Google have almost no safeguards against use in autonomous weapons systems. In tests, the models readily flagged civilians as military targets, and a simple video game framing in the prompt was enough to override even belated refusals. The providers have promised fixes. The article Anthropic, OpenAI, and Google models readily marked medics, journalists, and surrendering soldiers as kamikaze drone…

Read at The Decoder

TechPowerUp TheLostSwede SI

(PR) ASUS Announces Availability of the ProArt GR1X

ASUS today announced the availability of the ProArt GR1X mini PC. Built on the NVIDIA RTX Spark superchip, ProArt GR1X is a Windows PC designed as an always-on home for local AI agents. It is able to run large models on-device, keep prompts and project files private, and work with the agent and creative tools people already use. ProArt GR1X can be specified with 128 GB and 64 GB of unified memory. Built for agentic AI, powered by RTX Spark AI agents work best when they can…

Read at TechPowerUp

The Guardian World Edward Helmore

‘Plume’ of decomposing garbage, gas prompts emergency health order for LA county residents

Landfill operators have been ordered to provide hotel accommodations and other care to area residents Officials in a Los Angeles county have ordered the operators of a landfill to provide hotel accommodations and other care to nearby residents after it exploded with decomposing garbage and methane gas. A noxious, dark geyser of material and smoke began to spew from the troubled Chiquita Canyon landfill on Monday afternoon, prompting officials to issue advisories for nearby…

Read at The Guardian World

Axios Ben Geman

U.S. drivers are just absorbing high gasoline prices

Pain at the pump may have changed some votes, but it's apparently not prompting big changes in how much Americans drive. Why it matters: Most driving isn't especially optional, so people are absorbing the cost. This year Americans have paid an additional $88 billion for gasoline relative to a no-war baseline, according to Brown University's Climate Solutions Lab. The Energy Policy Research Foundation finds that U.S. households paid between $300 and $730 more on gasoline in…

Read at Axios

CBS News Politics Geopolitics

AI platforms planned U.S. terror attack when asked to

The incoming chair of the Senate Homeland Security Committee is preparing hearings on AI after government watchdogs found the four most-used platforms all created proposals for a U.S. terrorist attack when prompted by investigators.

Read at CBS News Politics

arXiv (research papers) Lipeng He Security

PIPETTE: Prompt Injection Purple-Teaming for Robust LLM Agents

Large language model (LLM) agents routinely interact with untrusted external data, exposing them to indirect prompt injection (IPI) attacks that hijack agents into executing attacker-chosen tasks. To uncover these vulnerabilities at scale, automated red-teaming methods often leverage reinforcement learning (RL) to generate attacks that achieve high attack success rates (ASR). Model developers use the attacks discovered during internal red-teaming to harden agents through…

Read at arXiv (research papers)

arXiv (research papers) Guolong Zheng SI

Semantic Context Handoff with On-Demand Evidence for Long-Horizon Agents

Traditional long-horizon LLM agents often hand off context through full-history prompting, retaining all previous action--observation interactions. As execution proceeds, this history keeps growing, increasing inference cost and potentially degrading accuracy. Existing methods control this growth by preserving compact conclusions while removing earlier observations, but doing so can discard evidence that later becomes relevant. We introduce Semantic Context Handoff (SCH)…

Read at arXiv (research papers)

Infosecurity Magazine Security

Java4Shell Vulnerability Allows Remote Attacks on Major Java Apps

A critical OpenJDK vulnerability could expose major Java applications to remote code execution, prompting CISA to urge organizations implement mitigations

Read at Infosecurity Magazine

arXiv (research papers) Stavros Damianakis Security

AutoPI: Taxonomy-Guided Adaptive Red Teaming for Prompt Injection in LLM Agents

Prompt injection is the dominant unsolved security risk for LLM agents, and the community now agrees that robustness must be measured against adaptive attackers rather than fixed benchmark injections. Many adaptive red teamers exist, and most are built to maximize attack success rate (ASR). Yet ASR counts how often a target breaks, not in how many different ways. A red teamer can achieve a high ASR by reusing a few techniques that work, leaving the others untested and absent…

Read at arXiv (research papers)

Iran International

Killing of Iranian schoolgirl sparks calls for legal reform

The killing of a 16-year-old Iranian girl, allegedly by her father, has prompted lawyers and women's rights activists to demand tougher penalties for child killings and greater accountability from schools and child protection agencies in the western city of Kermanshah.

Read at Iran International

The Verge Jess Weatherbed SI

Suno now lets you make AI videos for your AI music

Suno has introduced image and video generation in its latest effort to expand beyond being an AI music platform. The new feature, called "Visuals," allows users to turn Suno songs into AI-generated video clips that automatically lipsync to lyrics, or make images via prompt descriptions that can be used as thumbnails and album artwork. Visuals

Read at The Verge

The New York Times Asia Paul Sonne and Damien Cave

Putin Arrives in Asia With a Show of Force: Nuclear Bombers

Russian military planes flew through the region, prompting Japan and South Korea to scramble jets in response, and drawing complaints from Taiwan.

Read at The New York Times Asia

arXiv (research papers) Haoyang Liu Security

What Survives the Shift from Text to Pixels? Conditional Protection against Prompt Injection in Image-Rendered Documents

Multimodal language models increasingly answer questions over external documents, exposing them to indirect prompt injection: commands planted in untrusted content to hijack the response. Rendering untrusted content as page images lowers attack success on average, leaving open which injected operations survive rendering, at what utility cost, and whether failed image attacks reflect unread commands or commands read but not followed. We pair text and page-image versions of…

Read at arXiv (research papers)

arXiv (research papers) Bonan Ruan Security

Guarding the Kingdom of Skill-Based Agents in the AI Dynasty of Computation: Control-Flow Integrity for Agent Skills

Control-flow integrity (CFI) confines a program to a control-flow graph fixed ahead of time. Procedural agent skills move this control flow into prompt text, and a large language model (LLM) agent decides at runtime whether each step runs in its reasoning, a tool call, or a script. The procedure's control skeleton survives, but its enforceability depends on where each step runs. Many violations, such as overwriting a file before validation, use only operations the skill…

Read at arXiv (research papers)

The Guardian US

Smoke spews from landfill in California after gas release – video

A release of gas at the Chiquita Canyon landfill in Castaic, California, sent smoke billowing into the air and prompted an evacuation, the Los Angeles county fire department said. The department said workers doing routine digging in an area that had not been worked on for some time hit a deposit of gases. Officials initially said a pipe had been struck. There was no fire, only smoke, and a hazardous materials team was sent in with air monitoring in place, according to local…

Read at The Guardian US

Al Jazeera Geopolitics

Myanmar-linked arsenic pollution prompts clean water scramble in Thailand

Thai authorities race to provide safe water to villagers after arsenic from Myanmar mines contaminates Mekong River.

Read at Al Jazeera

Amwaj.media Amwaj.media Staff

Beirut killing of Iraqi ex-MP ignites reckoning with ‘sectarian’ past

The story: Almost every possible culprit has been blamed for the killing in Beirut of Iraqi Sunni former MP Ahmed Al-Alwani. Accusations range from Lebanese Hezbollah and Iran-aligned armed groups to actors from Syria. Lebanese investigators are seeking two Syrian suspects who reportedly fled across the border. In Iraq, the once-jailed politician has proved as divisive in death as in life, prompting Baghdad to threaten legal action over sectarian incitement. The coverage…

Read at Amwaj.media

ReliefWeb International Federation of Red Cross and Red Crescent Societies

South Sudan MVD Readiness - DREF Final Report (MDRSS018)

Countries: South Sudan, Ethiopia Source: International Federation of Red Cross and Red Crescent Societies Please refer to the attached file. Description of the Event Date when the trigger was met 23-12-2025 What happened, where and when? The Marburg Virus Disease (MVD) outbreak was declared in Ethiopia on 14 November 2025, prompting heightened preparedness measures in neighboring South Sudan due to the risk of cross-border transmission through frequent population movement…

Read at ReliefWeb

The Guardian World Jasper Jolly, Jillian Ambrose and Kiran Stacey

UK will miss climate targets if it weakens EV sales rules, advisers warn

Analysts say mooted move to let more petrol and diesel vehicles be sold would wreck plan to reach net zero by 2050 Weakening the UK’s electric car sales rules would probably cause the government to miss legally binding climate targets, according to advisers tracking Britain’s progress to net zero. The analysis by the Climate Change Committee (CCC) suggests the policy would blow a hole in the plan for achieving net zero carbon emissions by 2050 – prompting concerns that Andy…

Read at The Guardian World

Simon Willison SI

External sandboxes for coding agents

External sandboxes for coding agents I've been exploring new ways to sandbox code that is executed by agents, with a focus on making it safer to try out local models without worrying that a mistake or prompt injection attack might cause them to take damaging actions on a machine. I worked on this project with Prime Radiant, the AI research lab behind Superpowers and other projects. My post on the Prime Radiant blog introduces the external sandbox pattern, where code executed…

Read at Simon Willison

NBC News Doha Madani

Gas geyser and noxious odor from L.A. area landfill prompts public health response

Pungent odors wafting from the Chiquita Canyon Landfill prompted Los Angeles County officials to relocate residents after drilling unleashed a geyser of trash and fumes.

Read at NBC News

AWS Machine Learning Blog Dmitry Soldatkin SI

Amazon SageMaker AI real-time endpoints now support KV cache aware routing

Amazon SageMaker AI real-time endpoints now support KV cache-aware routing, a new strategy that routes each request to the instance holding the longest cached match for its prompt. Learn how it works, when to use it compared with prefix-aware routing, and how to enable, invoke, and verify it on single model and inference component endpoints.

Read at AWS Machine Learning Blog

Older stories