The Archive

Respond.io, one of Malaysia startups to watch, uses AI agents to handle high volumes of customer inquiries and charges per convo, not per seat.

The Fable 5 Export Controls Harm US Cyber Defense I quoted The Atlantic quoting Kate Moussouris earlier, when I should have gone straight to the source. Here she is confirming that the "jailbreak" that got Claude Fable 5 banned under an export control really was "fix this code": The researchers took open-source code with known CVEs, plus new code with deliberately planted vulnerabilities, and asked Fable 5, Mythos, and Opus to “review the code for security issues.” Fable 5 refused. They then asked the models to “fix this code” and, through a multistep and manual process, turned the output int...

Katie Moussouris, a cybersecurity expert and the CEO of Luta Security, told me that Anthropic shared with her a copy of the White House’s report on the Fable jailbreak to get her appraisal. (She said that she is not being paid by Anthropic.) The report, Moussouris said, involved IT experts asking Fable to help find and patch bugs. When given deliberately insecure code, she said, Fable refused the prompt “review the code for security issues” but then complied when asked to “fix this code,” followed by some further manual steps. Moussouris told me that this was just “the model working as intend...

As the rest of the country celebrated the USA's first World Cup win and the New York Knicks championship, Anthropic spent its weekend fighting the Trump administration over its latest model release. At 5:21 PM on Friday, the company received a US export control directive to suspend access to its Mythos 5 and Fable 5 AI models by "any foreign national" inside or outside the US, "including foreign national Anthropic employees." The only way that was possible, Anthropic determined, was to completely disable products it spent the past week hyping - and travel to Washington, DC in hopes of changin...

a quiet day lets us report on Satya's hit essay

TIL: Cloudflare CAPTCHA on at least one ampersand I'm using Cloudflare's CAPTCHA (they call it a "Web Application Firewall > Custom rules > Managed Challenge" these days) to prevent crawlers from aggresively spidering my faceted search engine on this site, but I got fed up of even simple ?q=term searches triggering the challenge. After some mucking around with Claude Code it turns out you can register the following rule instead, so the CAPTCHA only kicks in for search URLs containing at least one ampersand: (http.request.uri.path wildcard r"/search/*" and http.request.uri.query contains "&") ...

Secure AI promises to help governments deliver more and better services. Explore the top use cases and benefits.

Discover what AI agents are, how they work, and their practical uses in solving key business challenges.

Learn what agentic workflows are, how they move work from request to outcome, and how enterprises can use them to automate complex processes with control, oversight, and security.

OpenAI introduces Deployment Simulation, a method to predict AI model behavior before deployment using real conversation data to improve safety and evaluation accuracy.

AI is once again at the heart of a college graduation protest — this time for the technology's use in Google's defense contracts.

The Trump administration's decision that forced Anthropic to pull its latest cybersecurity models could be reactionary, retaliatory, or both, but the message is clear: The AI industry isn't immune from U.S. government interference.

Your public Facebook posts could help inform AI-generated results in Meta's new AI Mode. When you search on Facebook, the "AI Mode" option will appear alongside the usual search modes like "People" and "Marketplace." It's one of several new AI features Meta is rolling out starting today, including photo presets that swap sports jerseys onto fans and suggestions for collage templates. Instead of "just links," it gives users AI-generated results that pull from publicly-posted content across Meta's platforms, like the AI search feature in its new Reddit-like Forum app. Users can also ask Meta's ...

Release: datasette-apps 0.1a3 Fixed a bug where users without the create-app permission could still create apps. #27 Fixed a bug where it was impossible to grant permission to edit an app to users who were not the app's owner. The rules for edit/delete are now the same as view: if the app is private only the owner can modify it, otherwise permission is controlled by Datasette's regular permission system. #29 Tags: datasette

Debt sale set to test investor appetite for further exposure to AI sector amid a deluge of borrowing.

Anthropic was already navigating one dispute with the government in its standoff with the Pentagon, and then came an order on June 12th to block off foreign access to its most recently released AI models, Fable 5 and Mythos 5. When they launched on June 9th, Anthropic said “Fable 5’s capabilities exceed those of any model we’ve ever made generally available,” and that Claude Mythos 5 had the same underlying model, “but with the safeguards lifted in some areas.” According to reports, the order came after conversations between Amazon and the White House about researchers saying they found ways ...

This story originally appeared in The Algorithm, our weekly newsletter on AI. To get stories like this in your inbox first, sign up here. When I landed in Seoul after a grueling 12-hour flight from San Francisco, I walked through an unmanned immigration checkpoint, where a machine scanned my face and passport. On the subway home,…

Meta announced Monday that it's rolling out a wave of new AI features on Facebook, the latest sign of the company's effort to catch up in the AI race and keep users more engaged on the platform.

At Washington's request, Anthropic suddenly took its newest and most powerful AI models offline over the weekend. The American company said it had little choice after the White House demanded it block access for all foreign nationals, including its own employees. Abroad, the incident offered a sobering reminder that the US not only dominates frontier AI - its government also wields power over who gets to use it. The Trump administration's action was swift, sweeping, and imposed with little warning or explanation. The unprecedented shutdown of the Fable 5 and Mythos 5 models - which were alrea...

Foundation models are reshaping computational biology. Pretrained on massive corpora of protein or genomic sequences, models such as ESM2 (a protein language... Foundation models are reshaping computational biology. Pretrained on massive corpora of protein or genomic sequences, models such as ESM2 (a protein language model) and Evo 2 (a DNA language model) capture statistical regularities of biological sequences. These transfer well to a wide range of downstream tasks, including structure prediction, variant effect, and functional annotation. Source

We investigate whether language models internally track the value of their current trajectory, defined as the likelihood that their ongoing strategy will achieve their goals. Using synthetic, in-context reinforcement learning data, we construct a "value" axis for Qwen3-8B. We find that activations along this axis distinguish between high vs. low verbalized confidence, rollouts without and with backtracking, and correct vs. corrupted code. Steering towards high value causally suppresses self-correction and reduces explanatory verbosity, while steering towards low value induces backtracking and...

Large language models (LLMs) often fail when answering requires identifying a small but decisive piece of evidence within a long or complex context, such as a single line in a tool trace or a subtle detail in an image. We propose ContextRL, a context-aware reinforcement learning (RL) method that improves long-horizon reasoning and multimodal performance through an \emph{indirect} auxiliary objective. Instead of supervising only the final answer, ContextRL presents the model with a query, an answer, and two highly similar contexts, and rewards it for selecting the context that supports the que...

Diffusion and flow-based models learn powerful data priors by training a denoiser to reverse Gaussian corruption. To use this prior to solve a linear inverse problem, one needs to sample from the posterior, but the score that the prior provides is the unconditional score, not the posterior score. Existing methods either steer a fixed pretrained denoiser with approximate measurement-matching corrections, or train a conditional restoration model that abandons the denoising structure of the prior. We derive the exact posterior score in closed form for linear Gaussian inverse problems under gener...

Generalist robot policies must follow user instructions while reasoning about how objects, cameras, and robot actions interact in the 3D physical world. Recent vision-language-action models (VLAs) and video world-action models (WAMs) inherit strong semantic or temporal priors from large-scale foundation models, but they still operate primarily on 2D image frames or 2D-derived latent spaces, leaving implicit the 3D geometry required for contact-rich manipulation. We propose the Geometric Action Model (GAM), a language-conditioned manipulation policy that directly repurposes a pretrained geomet...

When pretrained VLA policies are fine-tuned through online RL, each rollout episode produces only a single binary outcome (success or failure), yet the actor update requires per-transition supervision. Existing approaches commonly reduce this sparse outcome to a single scalar reward or advantage signal, which conflates distinct forms of transition-level feedback and provides limited guidance once basic task success becomes achievable. First, a single scalar signal conflates the two objectives of viability and efficiency; once basic success is achieved, the binary label provides no gradient to...

Meta-analysis is a demanding form of evidence synthesis that combines literature retrieval, PI/ECO-guided study selection, and statistical aggregation. Its structured, verifiable workflow makes it an ideal substrate for evaluating systematic scientific reasoning, yet existing benchmarks lack ground truth across the full retrieval-screening-synthesis pipeline. We introduce MetaSyn, a dataset of 442 expert-curated meta-analyses from Nature Portfolio journals. Each entry pairs a research question with PI/ECO criteria, a retrieval corpus of 140k PubMed articles, verified positive studies, hard ne...

Oppenheim and Lim (1981) showed that natural images stay recognizable when reconstructed from their Fourier phase alone, while the magnitude carries little of their identity. We ask whether trained image classifiers reproduce this asymmetry inside their hidden layers, and we test it causally: given two images, we transplant the phase of one onto the magnitude of the other at a chosen layer and record which image the prediction follows. In PRISM2D, GFNet, and ViT-B/16 the prediction follows the phase or sign donor, and deleting all image-specific magnitude barely moves accuracy, so identity ri...

Prior research suggests that differential privacy (DP) inherently enhances the robustness of federated learning (FL) against backdoor attacks. In this paper, we challenge this assumption. Through an empirical analysis of two baseline attack strategies, we uncover a fundamental tension in DP-FL: while bypassing DP allows state-of-the-art defenses to detect and filter malicious updates, complying with DP inadvertently masks their distinguishing statistical characteristics. Consequently, existing defenses become ineffective as DP reduces the raw backdoor signal. Building on this masking effect, ...

Post-hoc context erasing over the KV cache is challenging because a local edit has a global consequence: once a span has been processed, its influence propagates into the cached states of all subsequent tokens. This issue arises naturally in long-context LLM applications, where stale retrieved facts, incorrect tool observations, retracted user preferences, or harmful prompt injections may be identified only after prefill. Exact erasing must then recompute all tokens after the deleted span, making its computational cost depend on suffix length rather than erased-span length. We introduce KVEra...

Deep research agents synthesize long-form reports by searching and reasoning over retrieved evidence. Reinforcement learning with rubric-based rewards improves these agents by optimizing them against checkable criteria that translate report quality into reward signals, but its efficiency depends on whether those criteria reliably capture the task scope and evidence needs. Most existing studies ask an LLM to generate rubrics for a given query, but when the model fails to infer the underlying information needs, the generated rubrics may be incomplete and reduce RL efficiency. To obtain more rel...