Apple CEO presses software chief for ambitious Siri AI strategy
Apple wants to turn Siri into an autonomous AI agent that can act independently for users.
Stories per day · 30 days
3 RSS
Relatedvlm2apple1captcha1claude1conditional-prediction1cti1dark-web1decision-making1defense1edr1ios1memory1open-weights1poisoning1pretraining1prompt-injection1rag1siri1
18 stories, newest first · one card per story
Stories tagged #agent
MemoryDojo benchmarks poisoning attacks on LLM agent memory
Benchmark with 155 trajectories demonstrates all tested memory systems are susceptible to poisoning.
LLM agents show authorization drift in multi-step tool use
Frontier LLMs exhibit drift rates of 0.43-0.61 when pursuing goals without adversaries in the Authorization Drift Benchmark.
Anthropic adds EDR support for Claude Cowork AI agents
Claude Desktop beta now allows EDR sensors in local VMs for Windows and macOS users.
StateFence Defends Blue-Team LLM Agents
StateFence preserves task state in SOC LLM agents against evidence-chain contamination from indirect prompt injections.
DPDM Improves Agent Decision Making via Shared Predictor
DPDM raises mean accuracy from 78.80% to 84.92% across 50 benchmarks using one frozen predictor.
DarKnight: Retrieval-Guided Agentic Framework for Breaking Dark Web CAPTCHA Gates
DarKnight uses local open-weight vision models to solve dark-web CAPTCHAs for threat investigators.
It appears .agent and .agi are about to be the hot new domains
ICANN listed 1,615 new TLD applications, including bids by OpenAI and Meta for .agent and .agi.
Critical Progress DataDirect GenAI Flaw Lets Malicious OpenAPI Files Execute OS Commands
Progress warns CVE-2026-91140 lets malicious OpenAPI files run OS commands in DataDirect GenAI agents.
5
CaptchaArena: A Large-Scale, Fine-Grained Dataset for Training Computer-Use Agents on Interactive CAPTCHAs
CaptchaArena releases 50K verified interactive CAPTCHA trajectories and a 9B agent reaching 71.7 Pass@1.
Agent Name Collision Attacks in Multi-Agent Systems
A multi-agent system vulnerability is identified where remote names are used as routing identifiers, causing wrong-peer dispatch.
Issuer-Sovereign Agentic Payments
This paper describes a method to keep issuing bank control over agent payments.
SR-Fraud: An Outcome-Supervised Reflective LLM Agent Framework for Non-Stationary Payment Fraud Detection
SR-Fraud uses a reflective LLM agent to detect emerging payment-fraud bursts before labels mature.
Parallel cut research time and cost in half with GPT‑6 Astra
GPT-6 Astra boosts AI agent research efficiency, delivering high-quality results in half the time with 50% code cost reduction.
Muse, Meta's extraordinarily privileged AI assistant, has a serious 0-day
Meta's AI assistant Muse has a serious zero-day vulnerability allowing app and terminal commands to control its account token.
Rare Event Estimation via Iterative Unalignment
Research paper on estimating rare events from agent outputs, proposing a new importance sampling method with large efficiency gains.
Beyond Predictable Paths: Redefining AI Security Incident Reporting for Agents
This research paper identifies key elements and open questions for AI agent security incident reporting and addresses vulnerability generalization research.
Agents That Edit Documents: Measuring Agentic PDF Forgery Against a Non-Agentic Control
AI agents can forge financial documents with high success rates and low costs, though the threat is overestimated by a factor of two.