Artificial intelligence (AI) systems employ learning algorithms that adapt to their users and environment, with learning either pre-trained or allowed to adapt during deployment. Because AI can optimize its behavior, a unit's factory model behavior can diverge after release, often at the perceived expense of safety, reliability, and human controllability. Since the Industrial Revolution, trust has ultimately resided in regulatory systems set up by governments and standards bodies. Research into human interactions with autonomous machines demonstrates a shift in the locus of trust: we must trust non-deterministic systems such as AI to self-regulate, albeit within boundaries. This radical shift is one of the biggest issues facing the deployment of AI in the European region.

The European Commission (EC) has acknowledged the importance artificial intelligence (AI) plays in forming Europe's future, identifying AI as the most strategic technology of the 21st century.a With a recent proposal on a Regulation Laying Down Harmonised Rules on Artificial Intelligenceb (EU Regulatory Framework for AI), the EC aims at introducing the first comprehensive legal framework on AI, which will identify specific risks for AI, provide a collection of high-risk application domains, propose specific requirements that AI systems should meet when used in such domains, and define obligations for users and providers (U.S. regulatory development relating to AIc). What clearly emerges from these efforts is the need for an AI that behaves in a responsible way. A clear and globally accepted definition of responsibility for AI systems is still under development, but will likely include notions such as fairness, security and privacy, explain-ability, safety, and reproducibility. Although safety and reproducibility are fundamental issues in AI research and its industrial application, we will not cover them here since they are requirements in many areas of technology, therefore not specific to AI.

In her 2020 State of the Union Address,a President of the European Commission von der Leyen called for Europe to lead the way on digital in the areas of data and artificial intelligence (AI). Artificial intelligence, data and robotics (ADR) present an opportunity and a challenge for Europe, a chance to improve the competitiveness of the European public and private sectors, and a challenge to translate Europe's core AI, data, and robotics strengths into a global market advantage (see Figure 1). Challenges for adoption of AI, data, and robotics in Europe. Working together, the Big Data Value Association (BDVA), the Confederation of Laboratories for Artificial Intelligence Research in Europe (CLAIRE), the European Laboratory for Learning and Intelligent Systems (ELLIS), the European Association for Artificial Intelligence (EurAI), and the European Robotics Association (euRobotics) have founded the AI, Data and Robotics Association (Adra) in order to establish an effective European Partnership on AI, Data and Robotics with the European Commission. The objective is to strengthen European competitiveness, societal well-being, and environmental sustainability.

Welcome to the second Communications Regional Special Section spotlighting European countries and Israel. On a relatively small portion of the Earth, this region includes almost 50 countries with enormous cultural and socioeconomic diversity that is also reflected in the richness of its business structures and computer science research. The first Hot Topic article in this section illustrates the high overall share of European public research on a global scale, and further highlights significant differences within the region. We are happy to report the authors in this special section represent 15 countries throughout Europe plus Israel. An important goal emphasized by the European Union (E.U.) and many individual countries is to attain digital sovereignty of the private and public sectors, while further developing areas of traditional industrial and design strengths into the future.

Can technology affect human morality? This is not an esoteric test question from a college philosophy class, but a real, growing concern among leading technologists and thinkers. Technologies like social media, smartphones, and artificial intelligence can create moral issues at scale, and technology experts specifically and society generally are struggling to navigate these issues. On the one hand, technology can empower us with better information on the consequences of our actions, as when we use the Internet to research how to reduce our environmental footprint. In the past, such information may have been inaccessible or impossible to source, but today we can easily arm ourselves with data that helps us make choices we perceive to be more moral.

Federal agencies and officials utilizing artificial intelligence systems need to vigilantly monitor and control for systemic and racial biases included in machine learning technology, according to a new report from the National Institute of Standards and Technology. This recommendation comes from an extensive report on how organizations and enterprises, both private and public, can cultivate better trust in artificial intelligence. "Bias is neither new nor unique to AI and it is not possible to achieve zero risk of bias in an AI system," the report begins. "AI systems do not operate in isolation. They help people make decisions that directly affect other people's lives," said Reva Schwartz, the principal investigator for AI bias at NIST and one of the report's authors.

It took less than six hours for drug-developing AI to invent 40,000 potentially lethal molecules. Researchers put AI normally used to search for helpful drugs into a kind of "bad actor" mode to show how easily it could be abused at a biological arms control conference. All the researchers had to do was tweak their methodology to seek out, rather than weed out toxicity. The AI came up with tens of thousands of new substances, some of which are similar to VX, the most potent nerve agent ever developed. Shaken, they published their findings this month in the journal Nature Machine Intelligence.

Energy giant Saudi Aramco says its 2021 net profit soared by more than 120 percent due to higher crude oil prices, as global economic growth recovered from a pandemic induced downturn. The announcement came on Sunday hours after Yemen's Houthi rebels – against whom Saudi Arabia leads a military coalition – targeted several locations, including Aramco facilities, in cross-border armed drone attacks. Aramco, Saudi Arabia's cash cow, did not say if the attacks caused any damage. "Aramco's net income increased by 124 percent to $110bn in 2021, compared to $49bn in 2020," the company said in a statement. Aramco achieved a net income of $88.2bn in 2019 before the coronavirus pandemic hit global markets, resulting in huge losses for the oil and aviation sectors, among others.

WASHINGTON – In the first weeks of the first major European land war of the 21st century, the United States has sent tank-killing weapons to Ukrainian forces, but not fighter jets. It is equipping embattled Ukrainian troops with lightweight "kamikaze" attack drones, but not, at least in an obvious way, conducting an aggressive cyberwar to degrade Russia's technological advantage. The White House will commit no U.S. or NATO planes to the skies above Ukraine, a move U.S. officials fear could risk turning a regional war into a global conflagration, but it is providing Ukraine with missiles that could accomplish the same task of destroying Russian aircraft. Such is the tenuous balance the Biden administration has tried to maintain as it seeks to help Ukraine lock Russia in a quagmire without inciting a broader conflict with a nuclear-armed adversary or cutting off potential paths to de-escalation. Navigating this path has led to a tangle of decisions and sometimes tortured distinctions when it comes to what kinds of assistance Washington should provide, even as the situation on the ground evolves, pictures of dead civilians circulate around the globe, and President Volodymyr Zelenskyy of Ukraine pleads with Congress and President Joe Biden to do more to help.

Augmenting pre-trained language models with knowledge graphs (KGs) has achieved success on various commonsense reasoning tasks. However, for a given task instance, the KG, or certain parts of the KG, may not be useful. Although KG-augmented models often use attention to focus on specific KG components, the KG is still always used, and the attention mechanism is never explicitly taught which KG components should be used. Meanwhile, saliency methods can measure how much a KG feature (e.g., graph, node, path) influences the model to make the correct prediction, thus explaining which KG features are useful. This paper explores how saliency explanations can be used to improve KG-augmented models' performance. First, we propose to create coarse (Is the KG useful?) and fine (Which nodes/paths in the KG are useful?) saliency explanations. Second, to motivate saliency-based supervision, we analyze oracle KG-augmented models which directly use saliency explanations as extra inputs for guiding their attention. Third, we propose SalKG, a framework for KG-augmented models to learn from coarse and/or fine saliency explanations. Given saliency explanations created from a task's training set, SalKG jointly trains the model to predict the explanations, then solve the task by attending to KG features highlighted by the predicted explanations. On three commonsense QA benchmarks (CSQA, OBQA, CODAH) and a range of KG-augmented models, we show that SalKG can yield considerable performance gains -- up to 2.76% absolute improvement on CSQA.