The Kremlin has "strongly" rejected an assessment by five European countries that the Russian state killed jailed opposition leader Alexey Navalny by poisoning him. Navalny, President Vladimir Putin's fiercest domestic opponent for years, died in an Arctic prison colony on February 16, 2024 while serving a 19-year sentence for "extremism", a charge he and his supporters said was punishment for his opposition work. "We naturally do not accept such accusations. We consider them biased and baseless," Kremlin spokesman Dmitry Peskov told reporters during a daily briefing call on Monday. "In fact, we strongly reject them," he added.

Trustworthy evaluations of dangerous capabilities are increasingly crucial for determining whether an AI system is safe to deploy. One empirically demonstrated threat is sandbagging - the strategic underperformance on evaluations by AI models or their developers. A promising defense is to monitor a model's chain-of-thought (CoT) reasoning, as this could reveal its intentions and plans. In this work, we measure the ability of models to sandbag on dangerous capability evaluations against a CoT monitor by prompting them to sandbag while being either monitor-oblivious or monitor-aware. We show that both frontier models and small open-sourced models can covertly sandbag against CoT monitoring 0-shot without hints. However, they cannot yet do so reliably: they bypass the monitor 16-36% of the time when monitor-aware, conditioned on sandbagging successfully. We qualitatively analyzed the uncaught CoTs to understand why the monitor failed. We reveal a rich attack surface for CoT monitoring and contribute five covert sandbagging policies generated by models. These results inform potential failure modes of CoT monitoring and may help build more diverse sandbagging model organisms.

Artificial intelligence could be misused to design highly toxic chemical and biological weapons, warn scientists. The computer algorithms are a force for good, identifying new forms of antibiotics and medicine to combat Covid infection. But four researchers involved in AI-based drug discovery have now found that the technology could easily be manipulated to search for toxic nerve agents. The four were asked by the Swiss Federal Institute for Nuclear, Biological and Chemical Protection to look at whether AI could be used by those with ulterior motives – and their AI came up with 40,000 potentially toxic drugs in six hours. They have highlighted their concerns in the journal Nature Machine Intelligence.

It was only in preparation for the Spiez conference that the researchers started to explore how artificial intelligence could be used to design toxic molecules. Until now they have used a commercial de novo molecule generator, MegaSyn, which is guided by machine learning model predictions of bioactivity in order to find new therapeutic inhibitors of targets for human diseases. The generative model was designed to penalise predicted toxicity and to reward predicted target activity. As usual, therapeutic drugs are carefully tested for toxicity. If, for instance, a very promising new drug to treat hypertension has been developed, but it damages one of the important heart channels, it cannot be used since it will be too dangerous.

An artificial intelligence model was able to create 40,000 chemical weapons compounds in just six hours, after being given the task by researchers. A team of scientists were using AI to look for compounds that could be used to cure disease, and part of this involves filtering out any that could kill a human. As part of a conference on potentially negative implications of new technology, biotech startup Collaborations Pharmaceuticals, from Raleigh, North Carolina, 'flipped a switch' in its AI algorithm, and had it find the most lethal compounds. The team wanted to see just how quickly and easily an artificial intelligence algorithm could be abused, if it were set on a negative, rather than positive task. Once in'bad mode' the AI was able to invent thousands of new chemical combinations, many of which resembled the most dangerous nerve agents in use today, according to a report by The Verge.

Meghan Markle was the most Googled person in Britain this year, the tech giant has revealed. The American actress, who married Prince Harry in May, has seen more people look up her name than any other figure in the UK. This is the second time the former Suits star has soared to the top of the search engine's'top trending people' list. Meghan Markle was the most Googled person in Britain this year, the tech giant has revealed. This is the second time the former Suits star has soared to the top of the search engine's'top trending people' list Celebrity Big Brother contestant Roxanne Pallett came in second place.

Fluid-pumping proteins that are embedded into jelly-like membranes could be used to intercept toxic nerve agents and deliver an antidote at the same time. Exposure to organophosphate nerve agents – which include Novichok nerve agents – can disrupt muscle function and lead to convulsions or difficulty breathing. Researchers have now shown that an enzyme can catalyse the breakdown of such compounds, creating a nanometre-sized pump that can both devour the nerve agent and spew out an antidote. "The nerve agents are their food, or fuel," …

Crimes that involve chemical, biological, radiological or nuclear (CBRN) materials pose a deadly threat not just to the target of the attack but to innocent bystanders and police investigators. Often, these crimes may involve unusual circumstances or they are terrorist-related incidents, such as an assassination attempt or the sending of poisons through the mail. In the recent notorious case of poisoning in the UK city of Salisbury in March 2018, a number of first responders and innocent bystanders were treated in hospital after two victims of chemical poisoning were found unconscious on a park bench. One policeman who attended the scene became critically ill after apparent exposure to a suspected chemical weapon, said to be a special nerve agent called novichok. Police said a total of 21 people required medical care after the incident.

Russia may soon use robots to carry out AI-powered assassinations as part of a new cold war, a computer expert has claimed. Future attacks on exiles and dissidents could be untraceable thanks to the use of automatons, warns Dr Jeremy Straub at Dakota State University. Intelligent machines could be programmed to hack into food ordering systems, like those used at fast food restaurants, to poison their victims. This would have the benefit of killing a target, while appearing to the outside world to be an accidental allergic reaction. The claims come in the wake of the high profile attack on former Russian spy Sergei Skripal and his daughter Yulia, which involved the use of a deadly nerve agent. Prime Minister Theresa May has laid the blame at Moscow's door, expelling a number of prominent diplomats from Britain in the process.