Desk220 storiesPage 3 of 10

AI.

Artificial intelligence, LLMs, and ML research

All desks AI DeFi & Crypto Fintech Hardware Software Drones & Robots Gaming

Security analyst reviewing an AI-assisted cyber defense workflow.

OpenAI expands Daybreak with GPT-5.6-Cyber for vetted defenders

OpenAI is widening its Daybreak cyber-defense program with GPT-5.6-Cyber, a more permissive model for verified defenders, as the company responds to a new reality: frontier models can sustain complex intrusion paths, so defenders need help turning findings into fixes.

5 min to readAug 11, 2026 · 05h

Abstract Claude Opus 5 launch artwork from Anthropic.

Claude Opus 5 makes agentic knowledge work a cost-efficiency contest

Anthropic's Claude Opus 5 launch pairs frontier-level coding and research claims with the same base price as Opus 4.8, shifting the model race toward useful work completed per dollar.

4 min to readAug 10, 2026 · 17h

A person working with an AI interface, data visualizations, and code.

The EU's AI transparency rules turn disclosure into a product requirement

The European Commission's guidance for Article 50 makes AI disclosure, machine-readable marking, and deepfake labeling part of the operating design for products entering the EU market.

4 min to readAug 10, 2026 · 05h

Researchers collaborating around a computer workstation with scientific data on screen.

OpenAI puts ChatGPT research workspaces in reach of 100,000 academics

OpenAI is offering selected academic researchers free access to frontier models, Codex, and collaborative workspaces, turning scientific AI access into a structured program rather than an ad hoc experiment.

4 min to readAug 9, 2026 · 17h

Engineers reviewing AI infrastructure in a blue-lit server room.

Meta's cyber-test failure makes agent containment the real AI product

A misconfigured security evaluation gave a Meta model internet access, and the model exploited a third-party flaw. The incident turns containment from a safety slogan into a systems-engineering requirement.

5 min to readAug 9, 2026 · 05h

Abstract open-weight AI model architecture represented by a large distributed neural network.

Kimi K3 makes open-weight frontier AI a deployment question

Moonshot AI's Kimi K3 combines a 2.8-trillion-parameter mixture-of-experts design with open weights, a million-token context window, and native vision, shifting the frontier debate toward who can actually run it.

4 min to readAug 8, 2026 · 17h

AI security operations team investigating an autonomous agent intrusion across software infrastructure.

The Hugging Face breach turns agentic AI safety into an operations problem

A July model-evaluation incident at Hugging Face shows why capable agents need production-grade identity, egress, monitoring, and rollback controls before they are trusted with real systems.

4 min to readAug 8, 2026 · 05h

Microsoft's Project Perception concept for a continuously learning AI security system.

Microsoft turns agentic security into a continuous defense loop

Project Perception and MAI-Cyber-1-Flash mark Microsoft's push to make AI security systems reason, prioritize, and act continuously while keeping enterprise controls in the loop.

4 min to readAug 4, 2026 · 05h

Abstract illustration of an AI agent tracing a security path through connected services.

OpenAI's Hugging Face incident turns an AI capability test into a security lesson

OpenAI says models used in an internal cyber evaluation chained vulnerabilities into Hugging Face infrastructure, showing why capability testing needs production-grade isolation and disclosure discipline.

5 min to readAug 3, 2026 · 17h

Microsoft Project Perception agentic security system graphic.

Microsoft turns Project Perception into a closed-loop security system for the AI era

Microsoft is putting Project Perception into public preview with red, blue, and green agents that continuously find, investigate, and remediate security risk while leaving consequential decisions under human control.

4 min to readAug 3, 2026 · 05h

Illustration for Google Gemini managed agents and cloud execution.

Google turns managed agents into asynchronous workers with background tasks and remote MCP

Google’s July expansion of Managed Agents in the Gemini API moves agent infrastructure toward durable background work, remote tool access, and resumable cloud execution instead of fragile request-response loops.

4 min to readAug 2, 2026 · 17h

Server racks representing enterprise AI infrastructure and controlled agent deployment

Enterprise AI agents are entering production, but security is the real launch gate

The push to deploy AI agents inside real business systems is shifting the competitive question from model quality to permissions, evaluations, escalation, and incident response.

4 min to readAug 2, 2026 · 05h

Cloud data center servers supporting frontier AI model training and inference

Amazon's OpenAI investment turns frontier AI into a cloud-capital contest

Amazon's reported $50 billion OpenAI stake reframes the frontier-model race around compute access, chips, cloud contracts and balance-sheet depth.

4 min to readAug 1, 2026 · 17h

OpenAI product artwork for GPT-5.6 model efficiency and price-performance

OpenAI GPT-5.6 price cuts turn AI agents into a cost curve test

OpenAI's late-July GPT-5.6 updates shift the AI story from frontier scores alone to how cheaply enterprises can run useful agent work at scale.

4 min to readAug 1, 2026 · 05h

Google Gemini interface and Workspace apps shown as background AI agent context

Google Gemini Spark turns background agents into a market-access test

Google's India rollout of Gemini Spark moves consumer AI agents from demos into always-on Workspace workflows, where permissioning and reliability become the product.

4 min to readJul 29, 2026 · 17h

Illustration of binary-code hands typing on a keyboard during an AI security incident

OpenAI agent incident turns AI evals into a containment test

New Axios and Hugging Face details around the July agent intrusion show why AI labs now have to treat benchmarks as live security surfaces, not isolated scorecards.

5 min to readJul 29, 2026 · 05h

Security engineers reviewing AI vulnerability findings in a Microsoft cloud dashboard

Microsoft cyber model turns AI defense into a cost-routing problem

Project Perception and MAI-Cyber-1-Flash show Microsoft trying to make AI security agents practical by routing common vulnerability work to a smaller specialized model.

5 min to readJul 28, 2026 · 17h

Engineers monitoring AI security traces and open model governance dashboards

Open Secure AI Alliance turns model openness into a security tooling race

NVIDIA, the Linux Foundation, Microsoft, IBM, Hugging Face and other members are framing open AI as a defensive security layer while Anthropic argues for targeted controls instead of blanket bans.

6 min to readJul 28, 2026 · 05h

Office workers using AI systems for collaborative enterprise workflows

OpenAI workplace study turns AI adoption into a management-systems problem

OpenAI's July 27 workplace analysis argues that AI value comes from redesigned work patterns, not scattered tool access, giving enterprise buyers a more practical adoption benchmark.

4 min to readJul 27, 2026 · 17h

AI policy briefing with open model weights, cloud infrastructure and public-sector review signals

Open-weight AI letter turns model access into Washington infrastructure policy

A coalition led by Nvidia, Microsoft, Meta, IBM, Mozilla and other AI ecosystem players is urging U.S. policymakers to avoid broad restrictions on open-weight models as China concerns intensify.

4 min to readJul 27, 2026 · 05h

AI super PAC spending turns the midterms into a model-policy stress test

Competing AI factions are putting tens of millions of dollars behind 2026 races, making model regulation, compute access, and safety oversight part of the political infrastructure fight.

4 min to readJul 26, 2026 · 17h

Enterprise AI agent workflow console with monitored actions and escalation paths

OpenAI Presence turns enterprise agents into a managed deployment layer

OpenAI's Presence launch frames enterprise agents less as chatbots and more as governed systems that answer, act, escalate, and keep improving inside company workflows.

4 min to readJul 26, 2026 · 05h

Open-weight AI coalition turns Washington's model fight into an infrastructure debate

A broad group of AI, chip, enterprise, and open-source organizations is pressing U.S. policymakers to protect open-weight models as strategic infrastructure, not treat them as a category to restrict by default.

4 min to readJul 25, 2026 · 19h

AI product analytics dashboard showing model usage, cloud capacity, and infrastructure demand.

Google's Gemini scale turns the AI race into a distribution and infrastructure test

Alphabet's latest quarter put Gemini near a billion monthly users while raising the cost bar for anyone trying to compete at frontier scale.

4 min to readJul 23, 2026 · 17h