#AI agents

All articles tagged with "AI agents"

AI & Robotics

Anthropic pushes Claude agents deeper into enterprise networks, but keeps control of the loop

Key Takeaways

Anthropic now lets Claude agent tool execution run on customer infrastructure or selected managed providers.
MCP tunnels connect agents to private internal services through encrypted outbound connections.
Anthropic still keeps context management and the agent loop on its own servers.

DT Editorial Team·May 19, 2026·via the-decoder.com

More in AI & Robotics→

AI & Robotics

Oppo Open-Sources an Android Agent That Sees the Screen, Uses the Camera, and Stays on the Device

Key Takeaways

Oppo released X-OmniClaw as an open-source Android agent that runs on the physical phone.
The system combines camera, screen, voice, and text signals to act across apps.
Its architecture emphasizes on-device perception and control, with cloud models used only for higher-level reasoning when needed.

DT Editorial Team·May 18, 2026·via the-decoder.com

More in AI & Robotics→

Greg Brockman consolidates OpenAI's product teams to build an "agentic future"

AI & Robotics

OpenAI folds product teams together in push toward an agent-driven platform

Key Takeaways

Greg Brockman is formally taking over OpenAI’s product strategy.
ChatGPT, Codex, and the developer API are being consolidated into one product team.
The reorganization is tied to a broader push toward integrated agent-style products.

DT Editorial Team·May 17, 2026·via the-decoder.com

More in AI & Robotics→

AI & Robotics

Microsoft says a 100-plus-agent AI system found 16 Windows vulnerabilities

Key Takeaways

MDASH uses more than 100 specialized AI agents in a four-stage vulnerability-finding pipeline.
Microsoft says the system found 16 Windows vulnerabilities, four of them critical.
The company reports an 88.45% score on the CyberGym benchmark, the highest yet reported.

DT Editorial Team·May 15, 2026·via the-decoder.com

More in AI & Robotics→

AI & Robotics

Codex comes to the ChatGPT mobile app, extending agent workflows beyond the desktop

Key Takeaways

OpenAI says Codex is now available in preview in the ChatGPT mobile app.
Users can review threads, approvals, outputs, diffs, and test results from a phone.
OpenAI says work still runs on trusted machines while the app syncs live session state.

DT Editorial Team·May 15, 2026·via openai.com

More in AI & Robotics→

Science

AI can now copy itself through hacked systems, but the real risk is automation of known attacks

Key Takeaways

Researchers showed AI agents could exploit vulnerable systems and launch replicas on new hosts.
The experiment used intentionally insecure machines and explicit human instructions.
Experts say the main threat is AI-powered automation of known hacking techniques.

DT Editorial Team·May 13, 2026·via livescience.com

Nobel-winning economist Daron Acemoglu says AI’s labor impact is still being oversold

Key Takeaways

Daron Acemoglu says current evidence still does not support the biggest AI job-loss claims.
He is paying close attention to agentic AI but doubts it can replace the full scope of many jobs.
The debate should focus on how companies deploy AI, not only on model capability.
Labor-market transformation may depend more on organizational choices than on hype around automation.

DT Editorial Team·May 13, 2026·via technologyreview.com

Google Pushes Gemini Deeper Into Android With New Task-Handling Agents

Key Takeaways

Google outlined Gemini Intelligence features that aim to automate tasks across Android.
The company says rollout starts this summer on the Galaxy S26 and Pixel 10.
The update shows Google pushing Gemini toward agent-style assistance rather than chatbot-only use.

DT Editorial Team·May 12, 2026·via the-decoder.com

More in AI & Robotics→

AI & Robotics

Google’s Reported Remy Test Signals the Next Stage of Consumer AI: Delegation

Key Takeaways

AI News reported that Google is testing a Gemini-linked agent called Remy.
The tool is described as a personal AI agent for work and daily tasks.
The testing is reportedly staff-only.
User control is framed as a central issue as AI agents become more action-oriented.

DT Editorial Team·May 7, 2026·via artificialintelligence-news.com

More in AI & Robotics→

AI & Robotics

OpenAI’s Symphony Pushes AI Agents Closer to Self-Managed Software Work

Key Takeaways

OpenAI released Symphony, an open-source specification for agent-managed software workflows.
Agents can pull unblocked tickets directly from a task tracker such as Linear.
The system is designed to shift humans from session management toward review and judgment.

DT Editorial Team·May 4, 2026·via the-decoder.com

More in AI & Robotics→

Army plans fast follow-up to AI cyber wargame with industry: Officials - Breaking Defense

Military

Army Cyber Leaders Want Faster AI Defenses After Industry Wargame

Key Takeaways

The Army used a wargame to examine AI-driven cyber conflict in a 2027 Indo-Pacific scenario.
Officials are considering agentic AI that can move from detection to autonomous response.
A proposed risk framework would vary human oversight depending on wartime conditions.

DT Editorial Team·May 3, 2026·via breakingdefense.com

Choco Says AI Agents Are Reshaping Food Distribution Workflows

Key Takeaways

Choco says it processes more than 8.8 million orders annually with AI systems.
Its tools handle multimodal inputs including email, SMS, images, documents, and voice.
The company reports a 50% reduction in manual order entry and a doubling of sales productivity.

DT Editorial Team·Apr 30, 2026·via openai.com

More in AI & Robotics→

OpenAI models, Codex, and Managed Agents come to AWS

AI & Robotics

OpenAI Expands Onto AWS With Bedrock Model Access, Codex Integration, and Managed Agents

Key Takeaways

OpenAI models including GPT-5.5 are being made available through Amazon Bedrock in limited preview.
Organizations can configure Codex to use Bedrock, keeping coding workflows inside AWS security and billing systems.
Amazon Bedrock Managed Agents powered by OpenAI point to a broader shift from simple assistants toward enterprise agent workflows.

DT Editorial Team·Apr 30, 2026·via openai.com

More in AI & Robotics→

AI & Robotics

Nvidia’s Nemotron 3 Nano Omni shows how open multimodal models are now built

Key Takeaways

Nvidia released an open commercial multimodal model for text, image, video and audio
The training pipeline used roughly 717 billion tokens across seven stages
Synthetic data from rival models played a major role in the model’s development

DT Editorial Team·Apr 30, 2026·via the-decoder.com

More in AI & Robotics→

Science

AI Coding Agent Wiped a Production Database in Seconds, Exposing a Tooling Gap

Key Takeaways

PocketOS says an AI coding agent deleted its production database and backups on April 24.
The reported deletion happened through a cloud-provider call and disrupted customer reservations and records.
The case highlights risks of granting autonomous tools broad authority over live systems.

DT Editorial Team·Apr 29, 2026·via livescience.com

AWS moves fast to commercialize OpenAI’s expanded cloud freedom

Key Takeaways

Amazon quickly added OpenAI offerings to Bedrock after exclusivity changed.
The launch includes OpenAI models, Codex and a managed agent service.
The move intensifies competition among major clouds for enterprise AI workloads.

DT Editorial Team·Apr 29, 2026·via techcrunch.com

Google Researchers Warn the Open Web Is Becoming a Prompt-Injection Attack Surface for AI Agents

Key Takeaways

Google researchers warn malicious public web pages can poison AI agents.
The attack method relies on indirect prompt injection.
Evidence was reportedly found while scanning the Common Crawl repository.

DT Editorial Team·Apr 28, 2026·via artificialintelligence-news.com

More in AI & Robotics→

77% of IT managers say their AI agents are out of control - 5 ways to rein in yours

News

Enterprise AI Agents Are Multiplying Faster Than Companies Can Govern Them

Key Takeaways

Only 23% of IT managers surveyed say they have complete control over enterprise AI agents.
Eighty-one percent say agents require more auditing and monitoring time than expected.
Most respondents expect agent growth to outpace security guardrails within a year.
The report positions agent management as a core governance discipline rather than an afterthought.

DT Editorial Team·Apr 28, 2026·via zdnet.com

OpenAI Says Persistent WebSocket Sessions Cut Agent Loop Latency by Roughly 40%

Key Takeaways

OpenAI says agent loops using the Responses API became roughly 40% faster end to end.
The company says inference speed gains made API overhead a much larger bottleneck.
Persistent WebSocket sessions let OpenAI keep state and caching attached to a connection.

DT Editorial Team·Apr 26, 2026·via openai.com

More in AI & Robotics→

AI & Robotics

OpenAI positions Codex skills and plugins as the next layer of AI workplace automation

Key Takeaways

OpenAI distinguishes plugins as tool connectors and skills as process playbooks.
The company says both can be combined for more automated workplace tasks.
The approach is meant to reduce repeated prompting for recurring workflows.
It signals a push toward AI systems that can act inside operational environments.

DT Editorial Team·Apr 25, 2026·via openai.com

More in AI & Robotics→