Users experimenting with DeepSeek have seen the Chinese AI chatbot reply and then censor itself in real time, providing an arresting insight into its control of information and opinion. Users might expect censorship to happen behind closed doors, before any information is shared. But that does not seem to be the case in the tool that sent US technology stocks tumbling on Monday. DeepSeek, or the automated guardrails that appear to police its own freedom of "thought" and "speech", brazenly deletes uncomfortable points. Before the censor's cut comes, DeepSeek seems remarkably thoughtful. In Mexico, Guardian reader Salvador asked it on Tuesday if free speech was a legitimate right in China.

There was a lot of news last week. The Guardian's journalism is independent. We will earn a commission if you buy something through an affiliate link. Donald Trump, Sam Altman, Masayoshi Son and Larry Ellison announced a 500bn initiative to expand infrastructure supporting artificial intelligence dubbed Stargate. On its heels came a press release from Meta vowing to expand its capital expenditure to 65bn in the coming year to expand its data centers.

If you ask a historian what happened on June 4, 1989, you will learn that this was the date that hundreds of protesters were killed in the Tiananmen Square Massacre. However, if you ask China's controversial ChatGPT rival, DeepSeek, you are unlikely to get any response at all. During testing, MailOnline watched in real-time as answers to sensitive questions were scrubbed and replaced with evasive replies. On topics censored by the Chinese Communist Party (CCP), the bot briefly attempts to respond before wiping away its own answers in front of the user's eyes. DeepSeek actively censors topics related to protests such as the White Paper Movement and Hong Kong's pro-democratic Umbrella Movement.

We study long-run progress in artificial intelligence in a quantitative way. Many measures, including traditional ones such as patents and publications, machine learning benchmarks, and a new Aggregate State of the Art in ML (or ASOTA) Index we have constructed from these, show exponential growth at roughly constant rates over long periods. Production of patents and publications doubles every ten years, by contrast with the growth of computing resources driven by Moore's Law, roughly a doubling every two years. We argue that the input of AI researchers is also crucial and its contribution can be objectively estimated. Consequently, we give a simple argument that explains the 5:1 relation between these two rates. We then discuss the application of this argument to different output measures and compare our analyses with predictions based on machine learning scaling laws proposed in existing literature. Our quantitative framework facilitates understanding, predicting, and modulating the development of these important technologies.

This paper proposes a Right to AI, which asserts that individuals and communities should meaningfully participate in the development and governance of the AI systems that shape their lives. Motivated by the increasing deployment of AI in critical domains and inspired by Henri Lefebvre's concept of the Right to the City, we reconceptualize AI as a societal infrastructure, rather than merely a product of expert design. In this paper, we critically evaluate how generative agents, large-scale data extraction, and diverse cultural values bring new complexities to AI oversight. The paper proposes that grassroots participatory methodologies can mitigate biased outcomes and enhance social responsiveness. It asserts that data is socially produced and should be managed and owned collectively. Drawing on Sherry Arnstein's Ladder of Citizen Participation and analyzing nine case studies, the paper develops a four-tier model for the Right to AI that situates the current paradigm and envisions an aspirational future. It proposes recommendations for inclusive data ownership, transparent design processes, and stakeholder-driven oversight. We also discuss market-led and state-centric alternatives and argue that participatory approaches offer a better balance between technical efficiency and democratic legitimacy.

Three publicly-available LLM specifically designed for legal tasks have been implemented and shown that classification accuracy can benefit from training over legal corpora, but why and how? Here we use two publicly-available legal datasets, a simpler binary classification task of ``overruling'' texts, and a more elaborate multiple choice task identifying ``holding'' judicial decisions. We report on experiments contrasting the legal LLM and a generic BERT model for comparison, against both datasets. We use integrated gradient attribution techniques to impute ``causes'' of variation in the models' perfomance, and characterize them in terms of the tokenizations each use. We find that while all models can correctly classify some test examples from the casehold task, other examples can only be identified by only one, model, and attribution can be used to highlight the reasons for this. We find that differential behavior of the models' tokenizers accounts for most of the difference and analyze these differences in terms of the legal language they process. Frequency analysis of tokens generated by dataset texts, combined with use of known ``stop word'' lists, allow identification of tokens that are clear signifiers of legal topics.

An LLM is stable if it reaches the same conclusion when asked the identical question multiple times. We find leading LLMs like gpt-4o, claude-3.5, and gemini-1.5 are unstable when providing answers to hard legal questions, even when made as deterministic as possible by setting temperature to 0. We curate and release a novel dataset of 500 legal questions distilled from real cases, involving two parties, with facts, competing legal arguments, and the question of which party should prevail. When provided the exact same question, we observe that LLMs sometimes say one party should win, while other times saying the other party should win. This instability has implications for the increasing numbers of legal AI products, legal processes, and lawyers relying on these LLMs.

Large Language Models (LLMs) have raised increasing concerns about their misuse in generating hate speech. Among all the efforts to address this issue, hate speech detectors play a crucial role. However, the effectiveness of different detectors against LLM-generated hate speech remains largely unknown. In this paper, we propose HateBench, a framework for benchmarking hate speech detectors on LLM-generated hate speech. We first construct a hate speech dataset of 7,838 samples generated by six widely-used LLMs covering 34 identity groups, with meticulous annotations by three labelers. We then assess the effectiveness of eight representative hate speech detectors on the LLM-generated dataset. Our results show that while detectors are generally effective in identifying LLM-generated hate speech, their performance degrades with newer versions of LLMs. We also reveal the potential of LLM-driven hate campaigns, a new threat that LLMs bring to the field of hate speech detection. By leveraging advanced techniques like adversarial attacks and model stealing attacks, the adversary can intentionally evade the detector and automate hate campaigns online. The most potent adversarial attack achieves an attack success rate of 0.966, and its attack efficiency can be further improved by $13-21\times$ through model stealing attacks with acceptable attack performance. We hope our study can serve as a call to action for the research community and platform moderators to fortify defenses against these emerging threats.

This study introduces a method for automating the consolidation process in a legal context, a time-consuming task traditionally performed by legal professionals. We present a generative approach that processes legislative texts to automatically apply amendments. Our method employs light quantized generative model, fine-tuned with LoRA, to generate accurate and reliable amended texts. To the authors knowledge, this is the first time generative models are used on legislative text consolidation. Our dataset is publicly available on HuggingFace1. Experimental results demonstrate a significant improvement in efficiency, offering faster updates to legal documents. A full automated pipeline of legislative text consolidation can be done in a few hours, with a success rate of more than 63% on a difficult bill.

Large Language Models (LLMs) have recently gained attention due to their ability to understand and generate sophisticated human-like content. However, ensuring their safety is paramount as they might provide harmful and unsafe responses. Existing LLM testing frameworks address various safety-related concerns (e.g., drugs, terrorism, animal abuse) but often face challenges due to unbalanced and obsolete datasets. In this paper, we present ASTRAL, a tool that automates the generation and execution of test cases (i.e., prompts) for testing the safety of LLMs. First, we introduce a novel black-box coverage criterion to generate balanced and diverse unsafe test inputs across a diverse set of safety categories as well as linguistic writing characteristics (i.e., different style and persuasive writing techniques). Second, we propose an LLM-based approach that leverages Retrieval Augmented Generation (RAG), few-shot prompting strategies and web browsing to generate up-to-date test inputs. Lastly, similar to current LLM test automation techniques, we leverage LLMs as test oracles to distinguish between safe and unsafe test outputs, allowing a fully automated testing approach. We conduct an extensive evaluation on well-known LLMs, revealing the following key findings: i) GPT3.5 outperforms other LLMs when acting as the test oracle, accurately detecting unsafe responses, and even surpassing more recent LLMs (e.g., GPT-4), as well as LLMs that are specifically tailored to detect unsafe LLM outputs (e.g., LlamaGuard); ii) the results confirm that our approach can uncover nearly twice as many unsafe LLM behaviors with the same number of test inputs compared to currently used static datasets; and iii) our black-box coverage criterion combined with web browsing can effectively guide the LLM on generating up-to-date unsafe test inputs, significantly increasing the number of unsafe LLM behaviors.