Full archive

Latest AI News

Search reporting on AI markets, models, products, companies, policy, and culture.

Showing 1-18 of 33Results matching "Agent harness"

Today

1 story

1 hour agoTools

Tools

Oracle Publishes Agent Design That Keeps Credentials and Proof Outside the Model

Oracle’s proposed harness treats a model’s output as a request for action, leaving systems outside the model to authorize work, retain state and verify its result.

2 min

Yesterday

3 stories

13 hours agoInfrastructure

Benchmark Infrastructure

InferenceX Puts GB300 First for MiniMax M2.7, but Only on a Fixed Chat Test

The new ranking gives infrastructure teams a comparable speed and cost snapshot, but its single-turn design does not test the multi-step agent work MiniMax highlights for M2.7.

2 min

23 hours agoStartups

Funding Startups

HiddenLayer Raises $100M for Runtime Controls on Enterprise AI Agents

The startup is funding a security layer meant to observe and intervene as agents use enterprise tools. Its test is whether buyers will keep purchasing that layer independently of the platforms hosting their AI.

2 min

YesterdayTools

Launch Tools

Google Launches Fairwind to Give Selected Defenders AI Tools for Finding and Fixing Flaws

The program draws a line between broadly available code-security tooling and restricted access to Gemini 3.8 Flash Cyber, reflecting the dual-use risk of AI that can identify and repair software weaknesses.

3 min

Earlier this week

5 stories

YesterdayProducts

Launch Products

CrowdStrike’s SafeMind Pits Red and Blue AI Agents in an NVIDIA Infrastructure Twin

The new Falcon system combines CrowdStrike’s security data and agent harnesses with NVIDIA’s open Nemotron models. Early tests favor the optimized open pipeline on detection quality, but cover only one attack scenario family.

3 min

YesterdayModels

Benchmark Models

Grok 4.6 Clears 50% on Biosecurity Refusal and Routine Biology Tasks

The reported result rewards a difficult balance: recognizing hazards hidden inside ordinary-looking research requests without shutting down legitimate biological work. Its surveillance score shows that refusal performance is not the whole biosecurity test.

3 min

Aug 31, 2026Models

Benchmark Models

Qwen3.8-Max Leads Open-Weight Commerce Agent Bench by Two Passes, With Audit Limits

The benchmark checks whether agents leave commerce software and files in the required state, rather than simply narrate a workflow. But configuration choices and unavailable immutable result bundles limit what the narrow ranking can settle.

2 min

Aug 31, 2026Models

Open release Models

DeepSeek Releases 305B Vision Weights Under MIT License, but Local Use Needs Serious Hardware

The release gives developers the model files and serving references needed to run or adapt DeepSeek-V4-Flash-Vision-Exp independently. Its 168GB size means that freedom is aimed more at infrastructure-equipped teams than typical individual users.

3 min

Aug 31, 2026Products

Launch Products

Modellix Adds 28+ Models, Including $0 Options, to DeepSeek Harness Beta

The beta gives users of the rapidly adopted coding agent one model-selection layer and web-tool routing, while its free-call pitch still requires a Modellix account and funded balance to create an API key.

3 min

Last week

9 stories

Aug 30, 2026Products

Platform shift Products

OpenAI Maps a Third Era of Persistent AI Coworkers, With Access and Reliability Still in the Way

The proposed shift would turn AI from a prompt-by-prompt helper into a longer-running collaborator, but its usefulness depends on shared context, workplace-system access, and dependable controls.

3 min

Aug 28, 2026Tools

Benchmark Tools

Perplexity Search API Sweeps Artificial Analysis Test, but Medium Context Wins

The result gives developers a concrete retrieval-quality lead to test, while showing that sending an agent more page text is not automatically the best route to better answers.

3 min

Aug 28, 2026Models

Research Models

Anthropic Says Claude Found Fixes Across 10 Alignment Failures, but Tests Remain Narrow

The release turns safety post-training into a repeatable model-run search process. Its value now depends on whether those benchmark gains survive broader tests and later training.

3 min

Aug 28, 2026Models

Benchmark Models

Meta’s EvoHarness-RL Takes Qwen3-8B to 96.9% on ALFWorld, Near Claude Opus 4.5

The result is a strong benchmark showing for learned state management, not yet evidence that the framework improves the long-running enterprise workflows it is designed to target.

3 min

Aug 28, 2026Tools

Open release Tools

Visa’s VVAH Adds AI Patch Writing Across 11 Stages, Leaving Merge Decisions to Humans

The open-source harness now moves from detection through candidate remediation and adversarial validation. Its default can alter a repository’s working copy, while build, test, review, and merge controls remain outside the tool.

3 min

Aug 27, 2026Products

Launch Products

Harness Launches Agent-Ready Repository and AI Review for High-Volume Code Changes

The products pair high-volume repository operations with inherited permissions and merge-time checks, but Harness still places the final decision to ship with people.

2 min

Aug 26, 2026Infrastructure

Launch Infrastructure

AWS Publishes Two AgentCore Paths to Query Cross-Account Knowledge Bases

The useful constraint is not simply access across AWS accounts: generated answers require an assumed role, while many workloads may not need an agent at all.

3 min

Aug 25, 2026Tools

Enterprise adoption Tools

Ora Adopts Vercel’s Eve After Its Benchmark Found Fewer Steps and Lower Costs

The adoption turns Ora’s testing stack into its production stack, letting it trace its own eve-powered agents in the same environment used to inspect rival harnesses on customer websites.

3 min

Aug 25, 2026Models

Launch Models

DeepSeek V4 Pro Lands on Fireworks With $2.50 CyberGym Solves, but Kimi K3 Scores Higher

The provider’s results frame security-agent selection as a three-way choice among raw solve rate, cost per completed task, and whether a model will reliably execute the required tool workflow.

3 min

Load more stories15 more next