We are releasing a new suite of security benchmarks for LLMs, CYBERSECEVAL 3, to continue the conversation on empirically measuring LLM cybersecurity risks and capabilities. CYBERSECEVAL 3 assesses 8 different risks across two broad categories: risk to third parties, and risk to application developers and end users. Compared to previous work, we add new areas focused on offensive security capabilities: automated social engineering, scaling manual offensive cyber operations, and autonomous offensive cyber operations. In this paper we discuss applying these benchmarks to the Llama 3 models and a suite of contemporaneous state-of-the-art LLMs, enabling us to contextualize risks both with and without mitigations in place.

The Mars Science Helicopter (MSH) mission aims to deploy the next generation of unmanned helicopters on Mars, targeting landing sites in highly irregular terrain such as Valles Marineris, the largest canyons in the Solar system with elevation variances of up to 8000 meters. Unlike its predecessor, the Mars 2020 mission, which relied on a state estimation system assuming planar terrain, MSH requires a novel approach due to the complex topography of the landing site. This work introduces a novel range-visual-inertial odometry system tailored for the unique challenges of the MSH mission. Our system extends the state-of-the-art xVIO framework by fusing consistent range information with visual and inertial measurements, preventing metric scale drift in the absence of visual-inertial excitation (mono camera and constant velocity descent), and enabling landing on any terrain structure, without requiring any planar terrain assumption. Through extensive testing in image-based simulations using actual terrain structure and textures collected in Mars orbit, we demonstrate that our range-VIO approach estimates terrain-relative velocity meeting the stringent mission requirements, and outperforming existing methods.

Estimating collision probabilities between robots and environmental obstacles or other moving agents is crucial to ensure safety during path planning. This is an important building block of modern planning algorithms in many application scenarios such as autonomous driving, where noisy sensors perceive obstacles. While many approaches exist, they either provide too conservative estimates of the collision probabilities or are computationally intensive due to their sampling-based nature. To deal with these issues, we introduce Deep Collision Probability Fields, a neural-based approach for computing collision probabilities of arbitrary objects with arbitrary unimodal uncertainty distributions. Our approach relegates the computationally intensive estimation of collision probabilities via sampling at the training step, allowing for fast neural network inference of the constraints during planning. In extensive experiments, we show that Deep Collision Probability Fields can produce reasonably accurate collision probabilities (up to 10^{-3}) for planning and that our approach can be easily plugged into standard path planning approaches to plan safe paths on 2-D maps containing uncertain static and dynamic obstacles. Additional material, code, and videos are available at https://sites.google.com/view/ral-dcpf.

It is important for official statistics production to apply ML with statistical rigor, as it presents both opportunities and challenges. Although machine learning has enjoyed rapid technological advances in recent years, its application does not possess the methodological robustness necessary to produce high quality statistical results. In order to account for all sources of error in machine learning models, the Total Machine Learning Error (TMLE) is presented as a framework analogous to the Total Survey Error Model used in survey methodology. As a means of ensuring that ML models are both internally valid as well as externally valid, the TMLE model addresses issues such as representativeness and measurement errors. There are several case studies presented, illustrating the importance of applying more rigor to the application of machine learning in official statistics.

In late August 2023, Ilya Gambashidze was in a conference room at the office of Social Design Agency, a Russian IT company he founded that is based in Moscow, close to the world-renowned Moscow Conservatory. Gambashidze was relatively unknown in Russian politics at the time, but just a month earlier his name had appeared on a Council of the European Union's list of Russian nationals subjected to sanctions for playing a central role in a sprawling disinformation campaign against Ukraine. In the conference room, Gambashidze was laying out his plans for a new target: along with his colleagues, he began drafting what would become known as the "Good Old USA Project." The project was supposed to influence the outcome of the US presidential election in favor of former president Donald Trump, specifically targeting certain minorities, swing state residents, and online gamers, among others, in a scheme that included a full time team dedicated to the cause. On Wednesday, Gambashidze and his company were named by the Department of Justice among the architects of a disinformation campaign known as Doppelganger that has for the last two years been targeting Ukraine, and more recently, the US elections.

The UK government has signed the first international treaty on artificial intelligence in a move that aims to prevent misuses of the technology, such as spreading misinformation or using biased data to make decisions. Under the legally binding agreement, states must implement safeguards against any threats posed by AI to human rights, democracy and the rule of law. The treaty, called the framework convention on artificial intelligence, was drawn up by the Council of Europe, an international human rights organisation, and was signed on Thursday by the EU, UK, US and Israel. The justice secretary, Shabana Mahmood, said AI had the capacity to "radically improve" public services and "turbocharge" economic growth, but that it must be adopted without affecting basic human rights. "This convention is a major step to ensuring that these new technologies can be harnessed without eroding our oldest values, like human rights and the rule of law," she said.

John Lewis brings back'never knowingly undersold' Getty Images Retailer John Lewis is bringing back its "never knowingly undersold" price pledge from Monday, two years after abandoning it. It will also apply to online sales for the first time, whereas it previously only applied to in-store shopping, and will use AI to match the prices of 25 top retailers. The department store chain has been trying to win back customers after a tough few years that has seen it cut jobs and close several stores. It swung back to profit earlier this year, but is expected to continue shedding jobs as it seeks to revive its fortunes. The decision by John Lewis' new managing director Pete Ruis to restore the price pledge marks a change of direction from his predecessor.

Thomas VanCott compares his son Jake's experience with autism to life on a tightrope. Upset the delicate balance and Jake, 18, plunges into frustration, slapping himself and twisting his neck in seemingly painful ways. Like many families with children on the autism spectrum, Jake's parents sought treatments beyond traditional speech and behavioral therapies. One that seemed promising was magnetic e-resonance therapy, or MERT, a magnetic brain stimulation therapy trademarked in 2016 by a Newport Beach-based company called Wave Neuroscience. The company licensed MERT to private clinics across the country that offered it as a therapy for conditions including depression, PTSD and autism. Those clinics described MERT as a noninvasive innovation that could improve an autistic child's sleep, social skills and -- most attractive to the VanCott family -- speech. It was expensive -- 9,000 -- and not covered by insurance.

As national legislation on deepfake pornography crawls its way through Congress, states across the country are trying to take matters into their own hands. Thirty-nine states have introduced a hodgepodge of laws designed to deter the creation of nonconsensual deepfakes and punish those who make and share them. Earlier this year, Democratic congresswoman Alexandria Ocasio-Cortez, herself a victim of nonconsensual deepfakes, introduced the Disrupt Explicit Forged Images and Non-Consensual Edits Act, or Defiance Act. If passed, the bill would allow victims of deepfake pornography to sue as long as they could prove the deepfakes had been made without their consent. In June, Republican senator Ted Cruz introduced the Take It Down Act, which would require platforms to remove both revenge porn and nonconsensual deepfake porn.

An Israeli air attack on a car in the occupied West Bank city of Tubas has killed at least five Palestinians and left others wounded, local media and medical workers report. Israel said early on Thursday morning that military aircraft took part in "three different attacks" on Palestinian fighters who "posed a threat" to their forces in the Tubas region. Palestinian medics reported that five people were killed in an Israeli air strike targeting a car and two people were injured, one of whom was critical, according to the Palestinian state news agency, Wafa. The Reuters news agency later reported that the number of those killed in the drone attack had risen to at least six. According to Wafa, the attack was carried out by an armed drone, and Palestine Red Crescent Society ambulance crews brought the bodies of the five deceased men and two wounded people to the Tubas Turkish Government Hospital.