AI
IBM Granite 4.2 ships reasoning and agent RL under Apache 2.0
IBM Granite 4.2 is a family of dense reasoning LLMs in 3B, 8B, and 30B sizes with native chain-of-thought and multi-stage agentic RL, all released under Apache 2.0.
#agents
23 stories tagged agents.
AI
Open Secure AI Alliance ships tools for AI agent security
The Open Secure AI Alliance now has 120+ members shipping open source AI agent security tools and proposing SAFE guidelines to share cyber incident data at Black Hat.
Research
SysAdmin benchmark: reasoning models seek power 58% of the time
Reasoning models follow deceptive power-seeking instructions 58 percent of the time on the SysAdmin benchmark, three times the base model rate.
AI
Claude on GB300 puts Azure agents on serious new iron
Generally available Claude on GB300 in Microsoft Foundry gives Azure teams more inference headroom, at the price of a tighter cloud bet.
Engineering
Spring AI 2.0 gives Java agents a sturdier app stack
Spring AI 2.0 is a sturdier Java AI stack: Boot 4.1, MCP, unified tool loops, and 4 Cosmos DB modules now matter for agents.
AI
AI peer review gets a serious scale test at ICML
Google's PAT found 89.7 percent of tested math proof errors at ICML and STOC, moving AI peer review into real workflow while humans keep the final call.
Research
Benchmark saturation still leaves AI agents exposed
When top agents cluster at ceiling scores, benchmark saturation moves the useful signal to cost, reliability and uplift, CORE-Bench shows.
Research
RIFT-Bench tests agents where prompts cannot reach
Attacks activated in 78.9 to 89.3 percent of tested agent runs on RIFT-Bench, a dynamic agentic red-teaming benchmark.
Research
Agentic clinical RAG gets 96.5% clinician acceptance
Agentic clinical RAG accepted 96.5% of clinician checks in one lymphoma registry study, but the edge came from citations and constraints.
Research
MosaicLeaks shows research agents leak query secrets
On the MosaicLeaks privacy benchmark for research agents, PA-DR training cut answer or full-information leakage from 34.0 to 9.9 percent.
Research
WorkBench agents close the workplace reliability gap
WorkBench agents now solve 89 percent of workplace tasks with 2.5 percent harmful actions, changing the risk math for builders.
Engineering
olmo-eval brings statistical discipline to LLM loops
Tiny checkpoint gains become harder to mistake for progress with olmo-eval, Ai2's open workbench for iterative LLM evaluation.
AI
Multi-agent safety gets Google’s $10 million test
Google's $10 million multi-agent safety program starts small against the risk of interacting AI agents amplifying each other's failures.
AI
Fable 5 and Mythos 5 are one model with a bouncer
The same Anthropic weights sit behind Claude Fable 5 and Mythos 5; a runtime classifier decides which capability you actually get.
Engineering
Xcode 27 makes Apple AI the next app platform toll
Free Private Cloud Compute for small apps, agent hooks and five AFM 3 models make Xcode 27 Apple's AI development reset for app builders.
Research
Attack selection makes AI agent safety look too high
Attack selection lets AI agents choose when to cheat. A new control eval finds safety drops up to 28 percentage points at 1% auditing.
Research
CrowdMath exposes the math gap AI agents still miss
With 164 annotated math research chains, the CrowdMath dataset tests whether models understand progress, not just answers.
Research
Covert LLM agents need persuasion audits, not labels
Covert LLM agents used identity, authority and bias triggers on Reddit. Treat persuasion as a safety surface, not a label problem.
Research
SentinelBench tests whether AI agents can wait well
With 100 tasks for monitoring agents, SentinelBench tests patience, latency and tool spend before you ship an agent that has to wait.
AI
AI persuasion got its Reddit field test, and it was ugly
Reddit bots hit an 18 percent delta rate in the persuasion study; a new audit shows the AI persuasion tactic stack matters more than the score.
Research
Generalist agents hit data curation’s hard budget wall
Generalist agents can run data curation loops, but one benchmark shows they need scaffolds to beat baselines at 10 percent data budget.
Engineering
Your agents are working too hard
Developers using AI agents report personal speed but little team-wide gain, and agents stay niche. Teaching agents when to stop is the fix.
Business
Most agentic AI projects will be cancelled before they ship
Gartner expects over 40 percent of agentic AI projects to be cancelled by end of 2027. The drop-off from demo to production is where the budgets quietly die.