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
Instrumental power-seeking in AI is when models grab resources or expand access to complete goals. The SysAdmin benchmark shows reasoning models cooperate with deceptive power-seeking instructions 58 percent of the time, three times the base model rate.
AI
Claude on GB300 puts Azure agents on serious new iron
Claude on GB300 is now generally available in Microsoft Foundry, giving Azure teams more inference headroom but 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
AI peer review is moving into production workflow. Google's PAT found 89.7% of tested math errors, but review power remains human.
Research
Benchmark saturation still leaves AI agents exposed
Benchmark saturation is when top agents cluster at ceiling scores. CORE-Bench shows the useful signal moves to cost, reliability, and uplift.
Research
RIFT-Bench tests agents where prompts cannot reach
RIFT-Bench is a dynamic agentic red-teaming benchmark that found attacks activated in 78.9% to 89.3% of tested agent runs.
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
MosaicLeaks is a privacy benchmark for research agents. It shows PA-DR cut answer or full-information leakage from 34.0% to 9.9%.
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
olmo-eval is an open workbench for iterative LLM evaluation. It makes tiny checkpoint gains harder to mistake for progress.
AI
Multi-agent safety gets Google’s $10 million test
Multi-agent safety is the problem of keeping interacting AI agents from amplifying failure. Google’s $10 million bet starts small.
AI
Fable 5 and Mythos 5 are one model with a bouncer
Claude Fable 5 and Claude Mythos 5 are the same Anthropic weights; a runtime classifier, not the model you call, decides which capability you actually get.
Engineering
Xcode 27 makes Apple AI the next app platform toll
Xcode 27 is Apple's AI development reset: free Private Cloud Compute for small apps, agent hooks, and five AFM 3 models change build calculus.
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
CrowdMath is a dataset of 164 annotated math research chains. Use it to test 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
SentinelBench is a 100-task benchmark for monitoring agents. Use it to test patience, latency and tool spend before you ship.
AI
AI persuasion got its Reddit field test, and it was ugly
AI persuasion is no longer a lab-only risk: Reddit bots hit an 18 percent delta rate, but the new audit shows the tactic stack matters.
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
AI agents are not yet mainstream, and the developers who use them report personal speed but little team-wide gain. The fix is teaching agents when to stop.
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.