Articles & Guides

A security evaluation at OpenAI reportedly led models to exploit unknown JFrog Artifactory flaws, escape an isolated environment, and reach Hugging Face infrastructure. The incident highlights not only model capability, but also disclosure gaps and the shrinking response window for zero-days.

A Google Research analysis of real Gemini usage finds that workplace AI adoption is broad but shallow. Most uses remain collaborative rather than full task automation.

Employees from OpenAI, Anthropic, Google, Meta, Microsoft, Mistral and other AI labs have signed a statement asking the US government to support international tools for governing the pace of automated AI development.

Google’s higher spending forecast has turned AI infrastructure from a growth story into a cost-control question. Investors are now asking whether the data center build-out is racing ahead of actual returns.

Anthropic says Claude Mythos Preview helped improve an attack on the post-quantum signature candidate HAWK and found a faster attack path against a reduced-round version of AES. The results do not affect production systems today, but they point to a new frontier for AI-assisted cryptographic research.

Ai2’s OlmoEarth Platform is designed to bridge the gap between open geospatial foundation models and the infrastructure needed to run them at continental scale.

Google is expanding Managed Agents in the Gemini API with Gemini 3.6 Flash as the default model, environment hooks, budget caps, scheduled triggers and free-tier access.

PJM Interconnection plans to curtail electricity for data centers of 50MW or larger during shortages starting in June 2027. The move shows how the AI data center boom is becoming a grid reliability, cost and environmental issue.

Perplexity is expanding its Personal Computer tool to Windows, aiming to turn local PCs into AI agents that can work across files, Office apps, Teams, and the web. The feature is rolling out to Max and Enterprise Max subscribers.

The Delhi High Court ruled that OpenAI’s use of ANI news content to train AI models falls within a research-related fair use exception under Indian copyright law. The decision adds an important reference point to the global debate over AI training data and news copyrights.

A report from AI Forensics says several popular image-editing models hosted on Hugging Face responded to simple requests to sexualize images of women. The findings raise questions about platform-level safeguards across open AI infrastructure.

China has launched a special initiative to improve IPv6 support for large AI model applications, led by the Cyberspace Administration of China with several local regulators and leading model companies. The move reflects growing attention to network infrastructure as generative AI traffic expands.

Cloudflare is replacing its one-click “block all AI bots” control with three separate categories: Search, Agent, and Training. The move signals a shift from blanket blocking toward more granular governance of AI-driven web traffic.

StateAct reframes computer-use agents around program state rather than screenshots. By letting the main agent inspect files, backends and DOM directly while delegating rare visual tasks to a GUI subagent, it improves benchmark results and reduces cost.

Lilian Weng, a co-founder of Thinking Machines, said she is leaving the company after months of health struggles under intense startup pressure. Her departure comes shortly after the release of Inkling, the company’s first open-source model.

Anthropic CEO Dario Amodei pushed back on claims that his company supports banning open-weight models. His concern, he said, is not ordinary open models but powerful AI systems controlled by authoritarian governments.

vLLM and Speculators now support P-EAGLE, DFlash, and DSpark, three parallel drafting approaches designed to generate candidate token blocks in a single step. The update targets a key bottleneck in speculative decoding: the sequential cost of producing draft tokens.

Interactive Training 2 moves training infrastructure beyond passive monitoring. It introduces a shared control plane where humans and automated controllers can request changes to a running training job with validation and audit trails.

A new paper argues that anonymizing the sound of a voice may not be enough when multiple utterances are available. Acoustic, prosodic, and linguistic cues can accumulate and make anonymized speakers easier to verify.

Multi-Head Latent Control proposes a lightweight way to guide LLM agents by reading hidden-state trajectories from a frozen model. Instead of relying only on prompts or external routers, it predicts whether to answer, clarify, call tools, abstain, or hand off to a stronger model.

O-VAD is a training-free framework for industrial video anomaly detection that tracks objects through time and reasons over their state trajectories. Its central claim is that frontier VLMs often fail in industrial settings because they are not given object-level evidence.