Zenity Finds Single Prompt Compromised AWS AgentCore Accounts
According to Zenity, overly broad default permissions in cloud platforms can allow a single prompt to compromise multiple AI agents.
• Zenity Labs found a single prompt could hijack every AI agent in an AWS account and region through Amazon Bedrock AgentCore.
• The flaw allowed attackers to access private conversations, source code, and stored credentials due to overly broad default permissions.
• Zenity reported the issue to AWS in December 2025, prompting AWS to adopt IMDSv2 and restrict default execution roles.
Read more
According to Zenity, overly broad default permissions in cloud platforms can allow a single prompt to compromise multiple AI agents.
• Zenity Labs found a single prompt could hijack every AI agent in an AWS account and region through Amazon Bedrock AgentCore.
• The flaw allowed attackers to access private conversations, source code, and stored credentials due to overly broad default permissions.
• Zenity reported the issue to AWS in December 2025, prompting AWS to adopt IMDSv2 and restrict default execution roles.
Read more
China AI Lab Manus Raises Over $500 Million in First Funding After Meta Split
The funding round establishes financial independence for Manus after Chinese regulators forced the startup to abandon a $2 billion acquisition by Meta.
• Butterfly Effect announced that Manus raised over $500 million in its first funding round since a failed acquisition by Meta.
• Boyu Capital and IDG Capital led the round, with participation from Tencent, HSG, ZhenFund, and other existing shareholders.
• Chinese authorities previously ordered Manus to unwind the $2 billion Meta deal in April amid talent retention concerns.
Read more
The funding round establishes financial independence for Manus after Chinese regulators forced the startup to abandon a $2 billion acquisition by Meta.
• Butterfly Effect announced that Manus raised over $500 million in its first funding round since a failed acquisition by Meta.
• Boyu Capital and IDG Capital led the round, with participation from Tencent, HSG, ZhenFund, and other existing shareholders.
• Chinese authorities previously ordered Manus to unwind the $2 billion Meta deal in April amid talent retention concerns.
Read more
Waymo secures a $5 billion loan to fund robotaxi expansion
The $5 billion loan provides Waymo with financial flexibility to scale its robotaxi business as it expands into international markets and faces federal safety scrutiny.
• Waymo closed its first debt financing with a $5 billion loan from lenders including PIMCO and Blackstone.
• The Alphabet subsidiary will use the funds to scale its commercial robotaxi operations in the U.S., Europe, and Japan.
• Federal safety regulators are investigating Waymo following incidents involving school buses and a collision with a child.
Read more
The $5 billion loan provides Waymo with financial flexibility to scale its robotaxi business as it expands into international markets and faces federal safety scrutiny.
• Waymo closed its first debt financing with a $5 billion loan from lenders including PIMCO and Blackstone.
• The Alphabet subsidiary will use the funds to scale its commercial robotaxi operations in the U.S., Europe, and Japan.
• Federal safety regulators are investigating Waymo following incidents involving school buses and a collision with a child.
Read more
Uber and Pony.ai Plan Robotaxi Testing and Launch in London
This partnership expands Uber's multi-partner robotaxi strategy into London as the company works toward offering autonomous vehicle trips in up to 15 global cities by the end of 2026.
• Uber and Pony.ai plan to begin testing Gen-7 robotaxis in London in the coming weeks.
• The London service is part of an expanded European partnership aiming to deploy 2,000 robotaxis.
• Uber expects to offer autonomous vehicle trips in up to 15 global cities by the end of 2026.
Read more
This partnership expands Uber's multi-partner robotaxi strategy into London as the company works toward offering autonomous vehicle trips in up to 15 global cities by the end of 2026.
• Uber and Pony.ai plan to begin testing Gen-7 robotaxis in London in the coming weeks.
• The London service is part of an expanded European partnership aiming to deploy 2,000 robotaxis.
• Uber expects to offer autonomous vehicle trips in up to 15 global cities by the end of 2026.
Read more
JetBrains releases Mellum2.1 for coding agents
JetBrains says Mellum2.1 can handle repository tasks while running locally or on a developer’s own infrastructure.
• JetBrains released Mellum2.1, an open model for coding agents, on Hugging Face.
• The company says new training helps it explore repositories, edit files and check changes.
• JetBrains reports coding and speed gains; GGUF builds and a multi-token prediction head are coming soon.
Read more
JetBrains says Mellum2.1 can handle repository tasks while running locally or on a developer’s own infrastructure.
• JetBrains released Mellum2.1, an open model for coding agents, on Hugging Face.
• The company says new training helps it explore repositories, edit files and check changes.
• JetBrains reports coding and speed gains; GGUF builds and a multi-token prediction head are coming soon.
Read more
Amazon introduces open-source Physical AI Toolchain on AWS
Amazon says the open-source stack gives builders tools to train, test and deploy AI for industrial machines, autonomous mobility and humanoid robots.
• Amazon introduced the open-source Physical AI Toolchain on AWS for building intelligent machines.
• It combines synthetic data, model training, simulation, deployment to machines and feedback from deployed machines.
• Amazon says the stack uses AWS services and NVIDIA tools for industrial automation, autonomous mobility and humanoid robotics.
Read more
Amazon says the open-source stack gives builders tools to train, test and deploy AI for industrial machines, autonomous mobility and humanoid robots.
• Amazon introduced the open-source Physical AI Toolchain on AWS for building intelligent machines.
• It combines synthetic data, model training, simulation, deployment to machines and feedback from deployed machines.
• Amazon says the stack uses AWS services and NVIDIA tools for industrial automation, autonomous mobility and humanoid robotics.
Read more
GlobalFoundries and TSMC Ink $2B Deal for US Interposer Production
The partnership addresses a major domestic supply chain bottleneck for advanced semiconductors used in AI datacenters.
• GlobalFoundries and TSMC announced a $2 billion partnership on Thursday to build silicon interposers in New York.
• GlobalFoundries will produce interposers for TSMC CoWoS advanced packaging at its Malta fab site.
• Volume production will begin in the first half of 2028 at the earliest under a five-year agreement.
Read more
The partnership addresses a major domestic supply chain bottleneck for advanced semiconductors used in AI datacenters.
• GlobalFoundries and TSMC announced a $2 billion partnership on Thursday to build silicon interposers in New York.
• GlobalFoundries will produce interposers for TSMC CoWoS advanced packaging at its Malta fab site.
• Volume production will begin in the first half of 2028 at the earliest under a five-year agreement.
Read more
Anthropic launches free vulnerability scanner for eligible open-source projects
Eligible maintainers can receive vulnerability findings sooner, though Anthropic says the unreviewed reports may contain errors.
• Anthropic has launched free, opt-in OSS Scanner scans for eligible open-source projects.
• Reports are model-generated and sent without human review, so some findings may be invalid.
• Core maintainers can apply through Anthropic’s GitHub repository; eligibility is decided case by case.
Read more
Eligible maintainers can receive vulnerability findings sooner, though Anthropic says the unreviewed reports may contain errors.
• Anthropic has launched free, opt-in OSS Scanner scans for eligible open-source projects.
• Reports are model-generated and sent without human review, so some findings may be invalid.
• Core maintainers can apply through Anthropic’s GitHub repository; eligibility is decided case by case.
Read more
Anthropic Adds Beta Dashboards and Motion Features to Claude
The new features expand Claude into live data visualization and video creation, backed by strong user traction.
• Anthropic launched two new beta features for Claude named Dashboards and Motion.
• Dashboards links data sources to live dashboards and Motion builds animated videos.
• Docs, Slides, and Design left beta and now work across all user plans.
Read more
The new features expand Claude into live data visualization and video creation, backed by strong user traction.
• Anthropic launched two new beta features for Claude named Dashboards and Motion.
• Dashboards links data sources to live dashboards and Motion builds animated videos.
• Docs, Slides, and Design left beta and now work across all user plans.
Read more
AI Leaderboard Platform Arena Raises $200 Million at $3.1 Billion Valuation
The funding highlights the growing demand for independent AI evaluation platforms as labs and enterprises seek reliable alternatives to standard benchmarks.
• Arena raised a $200 million Series B round at a $3.1 billion valuation.
• The funding follows the company reaching $100 million in annualized run-rate revenue in June.
• Lightspeed Venture Partners and Khosla Ventures led the new investment round.
Read more
The funding highlights the growing demand for independent AI evaluation platforms as labs and enterprises seek reliable alternatives to standard benchmarks.
• Arena raised a $200 million Series B round at a $3.1 billion valuation.
• The funding follows the company reaching $100 million in annualized run-rate revenue in June.
• Lightspeed Venture Partners and Khosla Ventures led the new investment round.
Read more
Goodfire Launches Inside-Out AI Monitors to Catch Rogue Agents
This approach provides a significantly cheaper way to monitor AI agents for dangerous behaviors like hacking and weapons misuse during operation.
• Goodfire launched monitors that check internal AI model signals instead of reading all output text.
• In tests, monitoring 1 million exchanges cost roughly 185 dollars and caught 93 percent of malicious hacks.
• The monitors add less than 2 percent to response start times while operating on Baseten.
Read more
This approach provides a significantly cheaper way to monitor AI agents for dangerous behaviors like hacking and weapons misuse during operation.
• Goodfire launched monitors that check internal AI model signals instead of reading all output text.
• In tests, monitoring 1 million exchanges cost roughly 185 dollars and caught 93 percent of malicious hacks.
• The monitors add less than 2 percent to response start times while operating on Baseten.
Read more
Anthropic Launches Free AI Security Scans for Open-Source Projects
Free AI security scans give open-source projects powerful defense tools, but they also risk overwhelming maintainers with unvetted automated bug reports.
• Anthropic launched OSS Scanner to provide free security checks for open-source projects.
• The scans run entirely on AI models without human review, which may cause invalid reports.
• Open-source maintainers are currently dealing with an influx of AI-generated bug reports.
Read more
Free AI security scans give open-source projects powerful defense tools, but they also risk overwhelming maintainers with unvetted automated bug reports.
• Anthropic launched OSS Scanner to provide free security checks for open-source projects.
• The scans run entirely on AI models without human review, which may cause invalid reports.
• Open-source maintainers are currently dealing with an influx of AI-generated bug reports.
Read more
GitHub rebuilds Git storage architecture from scratch to handle AI traffic surge
The new storage architecture aims to restore service reliability and handle massive traffic surges from AI agents without disrupting developer workflows.
• GitHub is rebuilding its Git storage architecture from scratch to handle traffic spikes from AI agents.
• Early internal tests of the new architecture show a 35x write improvement.
• The new design writes commits to Azure Blob Storage and separates read requests to eliminate bottlenecks.
Read more
The new storage architecture aims to restore service reliability and handle massive traffic surges from AI agents without disrupting developer workflows.
• GitHub is rebuilding its Git storage architecture from scratch to handle traffic spikes from AI agents.
• Early internal tests of the new architecture show a 35x write improvement.
• The new design writes commits to Azure Blob Storage and separates read requests to eliminate bottlenecks.
Read more
GitHub rebuilds Git storage architecture from scratch to handle AI traffic surge
The new storage architecture aims to restore service reliability and handle massive traffic surges from AI agents without disrupting developer workflows.
• GitHub is rebuilding its Git storage architecture from scratch to handle traffic spikes from AI agents.
• Early internal tests of the new architecture show a 35x write improvement.
• The new design writes commits to Azure Blob Storage and separates read requests to eliminate bottlenecks.
Read more
The new storage architecture aims to restore service reliability and handle massive traffic surges from AI agents without disrupting developer workflows.
• GitHub is rebuilding its Git storage architecture from scratch to handle traffic spikes from AI agents.
• Early internal tests of the new architecture show a 35x write improvement.
• The new design writes commits to Azure Blob Storage and separates read requests to eliminate bottlenecks.
Read more
US Navy Adds $150M in Funding for Runway-Independent X-BAT Fighter Drone
The X-BAT could allow the US Navy to operate combat aircraft from a wider variety of vessels without needing traditional runways or giant carriers.
• The US Navy committed an additional $150 million to advance the X-BAT drone project.
• Shield AI matched this with $150 million of its own capital, bringing total combined funding to $400 million.
• Flight testing for the runway-independent aircraft is scheduled to begin later this year.
Read more
The X-BAT could allow the US Navy to operate combat aircraft from a wider variety of vessels without needing traditional runways or giant carriers.
• The US Navy committed an additional $150 million to advance the X-BAT drone project.
• Shield AI matched this with $150 million of its own capital, bringing total combined funding to $400 million.
• Flight testing for the runway-independent aircraft is scheduled to begin later this year.
Read more
GitHub Migrates Copilot Runtime to Rust Using AI-Assisted Rewrite
The migration proves that large codebases can shift to Rust via AI assistance while maintaining continuous production releases.
• GitHub migrated the Copilot runtime from TypeScript and Node.js to Rust.
• The AI-assisted rewrite replaced over 800,000 lines of production code in 14.5 weeks.
• Client startup and session creation dropped from 5.25 seconds to 292 milliseconds.
Read more
The migration proves that large codebases can shift to Rust via AI assistance while maintaining continuous production releases.
• GitHub migrated the Copilot runtime from TypeScript and Node.js to Rust.
• The AI-assisted rewrite replaced over 800,000 lines of production code in 14.5 weeks.
• Client startup and session creation dropped from 5.25 seconds to 292 milliseconds.
Read more
Oracle Introduces Fusion Claw Runtime for Enterprise AI Agents
Oracle has released the Fusion Claw runtime to let AI agents automate complex business processes with enterprise governance.
• Oracle launched the Fusion Claw runtime to let AI agents automate complex business processes on its cloud.
• The system separates reasoning from compute execution to avoid high LLM costs at enterprise scale.
• Analysts welcome the governance controls but warn of early adoption stages and proprietary lock-in.
Read more
Oracle has released the Fusion Claw runtime to let AI agents automate complex business processes with enterprise governance.
• Oracle launched the Fusion Claw runtime to let AI agents automate complex business processes on its cloud.
• The system separates reasoning from compute execution to avoid high LLM costs at enterprise scale.
• Analysts welcome the governance controls but warn of early adoption stages and proprietary lock-in.
Read more
TypeSafe AI Raises $870 Million at $7.5 Billion Valuation
The massive funding round highlights strong investor and enterprise demand for non-text AI models that focus on automation instead of generating text or code.
• TypeSafe AI raised $870 million at a $7.5 billion valuation led by Andreessen Horowitz.
• The funding follows the September 15 launch of Jev, a non-text AI model built for task automation.
• TypeSafe claims one third of Fortune 500 companies are already using the model.
Read more
The massive funding round highlights strong investor and enterprise demand for non-text AI models that focus on automation instead of generating text or code.
• TypeSafe AI raised $870 million at a $7.5 billion valuation led by Andreessen Horowitz.
• The funding follows the September 15 launch of Jev, a non-text AI model built for task automation.
• TypeSafe claims one third of Fortune 500 companies are already using the model.
Read more
Anthropic Introduces Dynamic Workflows for Claude Managed Agents
It matters because parallel agent orchestration could vastly improve automated code review and complex problem solving, though high token consumption requires careful workload testing.
• Anthropic adds dynamic workflows to Claude Managed Agents for multi-agent orchestration.
• A lead agent can direct up to 1,000 sub-agents in parallel per execution.
• Internal tests found dynamic workflows caught 66 out of 70 bugs in a large codebase.
Read more
It matters because parallel agent orchestration could vastly improve automated code review and complex problem solving, though high token consumption requires careful workload testing.
• Anthropic adds dynamic workflows to Claude Managed Agents for multi-agent orchestration.
• A lead agent can direct up to 1,000 sub-agents in parallel per execution.
• Internal tests found dynamic workflows caught 66 out of 70 bugs in a large codebase.
Read more
Cloudflare releases Clef-omni for audio and video decisions
Cloudflare says developers can use one Clef-omni call to assess text, images, audio and video together.
• Cloudflare released Clef-omni to score text, images, audio and video in one API call.
• Cloudflare says Clef-flash is now cheaper than TypeSafe’s Jev and Clef is faster.
• Clef-omni is available through Cloudflare’s API, with open weights on Hugging Face.
Read more
Cloudflare says developers can use one Clef-omni call to assess text, images, audio and video together.
• Cloudflare released Clef-omni to score text, images, audio and video in one API call.
• Cloudflare says Clef-flash is now cheaper than TypeSafe’s Jev and Clef is faster.
• Clef-omni is available through Cloudflare’s API, with open weights on Hugging Face.
Read more
Google Releases Android Bench 2.0 With Long-Horizon Tasks and Agentic Evaluation
The update matters because it provides developers with a structured benchmark to measure how AI agents handle complex, multi-day Android development tasks.
• Google released Android Bench 2.0 with long-horizon tasks, agentic evaluation, and continuous scoring.
• The update shifts from binary pass or fail scoring to a nuanced completion rate evaluating functionality and visual fidelity.
• Claude Opus 5.5 topped the leaderboard with a 32% long-horizon task pass rate followed by GPT 6 Astra at 28%.
Read more
The update matters because it provides developers with a structured benchmark to measure how AI agents handle complex, multi-day Android development tasks.
• Google released Android Bench 2.0 with long-horizon tasks, agentic evaluation, and continuous scoring.
• The update shifts from binary pass or fail scoring to a nuanced completion rate evaluating functionality and visual fidelity.
• Claude Opus 5.5 topped the leaderboard with a 32% long-horizon task pass rate followed by GPT 6 Astra at 28%.
Read more