Warning: This paper contains model outputs that are offensive in nature.

To this end, we propose evaluating the robustness of open-source large VLMs in the most realistic and high-risk setting, where adversaries have only black-box system access and seek to deceive the model into returning the targeted responses.

Google's Gemini 3 model keeps the AI hype train going - for now Google's latest model reportedly beats its rivals in several benchmark tests, but issues with reliability mean concerns remain over a possible AI bubble Gemini 3 is Google's latest AI model Google's latest chatbot, Gemini 3, has made significant leaps on a raft of benchmarks designed to measure AI progress, according to the company. These achievements may be enough to allay fears of an AI bubble bursting for the moment, but it is unclear how well these scores translate to real-world capabilities. What's more, persistent factual inaccuracies and hallucinations that have become a hallmark of all large language models show no signs of being ironed out, which could prove problematic for any uses where reliability is vital. AI may blunt our thinking skills - here's what you can do about it In a blog post announcing the new model, Google bosses Sundar Pichai, Demis Hassabis and Koray Kavukcuoglu write that Gemini 3 has "PhD-level reasoning", a phrase that competitor OpenAI also used when it announced its GPT-5 model . As evidence for this, they list scores on several tests designed to test "graduate-level" knowledge, such as Humanity's Last Exam, a set of 2500 research-level questions from maths, science and the humanities.

The essence of multi-modal fusion lies in exploiting the complementary information inherent in diverse modalities.

How generative AI in Arc Raiders started a scrap over the gaming industry's future Don't get Pushing Buttons delivered to your inbox? A rc Raiders is, by all accounts, a late game-of-the-year contender. Dropped into a multiplayer world overrun with hostile drones and military robots, every human player is at the mercy of the machines - and each other. Can you trust the other raider you've spotted on your way back to humanity's safe haven underground, or will they shoot you and take everything you've just scavenged? Perhaps surprisingly, humanity is (mostly) choosing to band together, according to most people I've talked to about this game.

Google's ultra-private CEO Sundar Pichai is showing me around Googleplex, its California headquarters. A walkway runs along the length of it, passing by a giant dinosaur skeleton, a beach volleyball pitch and dozens of Googlers lunching under the hazy November sun. But it's a laboratory, hidden away at the back of the campus behind some trees, that he is most excited to show me. This is where the invention that Google believes is its secret weapon is being developed. Known as a Tensor Processing Unit (or TPU), it looks like an unassuming little chip but, says Mr Pichai, it will one day power every AI query that goes through Google.

Former US treasury secretary Larry Summers is stepping down from the board at OpenAI, a week after a tranche of emails between him and late convicted sex offender Jeffrey Epstein was released. Summers said in a statement to the BBC that he was grateful for the opportunity to have served, excited about the potential of the company, and look forward to following their progress. Summers, who was also once the president of Harvard University, said on Monday that he would be stepping back from public commitments over his ties to Epstein. The recently released emails showed Summers communicated with Epstein until the day before Epstein's 2019 arrest for the alleged sex trafficking of minors. In a statement, the artificial intelligence company said it respected Summers' decision to resign.

A group of quantum physicists at Spanish firm Multiverse Computing claims to have created a version of the powerful reasoning AI model DeepSeek R1 that strips out the censorship built into the original by its Chinese creators. In China, AI companies are subject to rules and regulations meant to ensure that content output aligns with laws and "socialist values." As a result, companies build in layers of censorship when training the AI systems. When asked questions that are deemed "politically sensitive," the models often refuse to answer or provide talking points straight from state propaganda. Multiverse Computing specializes in quantum-inspired AI techniques, which it used to create DeepSeek R1 Slim, a model that is 55% smaller but performs almost as well as the original model. It allowed them to identify and remove Chinese censorship so that the model answered sensitive questions in much the same way as Western models.