Tesla starts Optimus production as third-quarter profit falls 43%
Tesla is investing in Optimus and AI while its third-quarter net income fell 43% and vehicle sales revenue remained flat.
• Tesla has started early Optimus robot production at its Fremont factory in California.
• Third-quarter net income fell 43% to $787 million as AI and Optimus spending weighed on profit.
• Tesla said some customer vehicles could join its robotaxi fleet around Q2 next year.
Read more
Tesla is investing in Optimus and AI while its third-quarter net income fell 43% and vehicle sales revenue remained flat.
• Tesla has started early Optimus robot production at its Fremont factory in California.
• Third-quarter net income fell 43% to $787 million as AI and Optimus spending weighed on profit.
• Tesla said some customer vehicles could join its robotaxi fleet around Q2 next year.
Read more
Tesla says it will start selling Cybercabs by year-end
If Tesla’s dates hold, fleet customers could buy Cybercabs before owners receive unsupervised FSD, while a federal investigation could affect their legal operation.
• Tesla says it will start selling Cybercabs by the end of the year, beginning on a small scale.
• Cybercabs produced today are going to Tesla’s own robotaxi fleet, according to its engineering vice president.
• NHTSA is investigating Tesla’s certification of the Cybercab, which has no steering wheel or other traditional controls.
Read more
If Tesla’s dates hold, fleet customers could buy Cybercabs before owners receive unsupervised FSD, while a federal investigation could affect their legal operation.
• Tesla says it will start selling Cybercabs by the end of the year, beginning on a small scale.
• Cybercabs produced today are going to Tesla’s own robotaxi fleet, according to its engineering vice president.
• NHTSA is investigating Tesla’s certification of the Cybercab, which has no steering wheel or other traditional controls.
Read more
1Password lets AI agents sign in without seeing passwords
The feature lets users approve agent sign-ins for tasks involving their accounts without giving the agents their passwords.
• 1Password for Agents lets AI agents sign in without seeing or handling passwords.
• Users approve sign-in requests on their own devices before 1Password fills in credentials.
• The system is designed for several AI agents, and 1Password says more partners are working on support.
Read more
The feature lets users approve agent sign-ins for tasks involving their accounts without giving the agents their passwords.
• 1Password for Agents lets AI agents sign in without seeing or handling passwords.
• Users approve sign-in requests on their own devices before 1Password fills in credentials.
• The system is designed for several AI agents, and 1Password says more partners are working on support.
Read more
OpenAI agents analyze mouse brain data in 32 minutes
OpenAI says the findings could help researchers choose which mechanisms to test in the lab first.
• Thousands of OpenAI agents analyzed data from millions of mouse cells in 32 minutes.
• The team estimates that the same analysis by hand would take several months to a year.
• In blind comparisons, experts preferred the distributed agents’ reports over a single agent’s 75 percent of the time.
Read more
OpenAI says the findings could help researchers choose which mechanisms to test in the lab first.
• Thousands of OpenAI agents analyzed data from millions of mouse cells in 32 minutes.
• The team estimates that the same analysis by hand would take several months to a year.
• In blind comparisons, experts preferred the distributed agents’ reports over a single agent’s 75 percent of the time.
Read more
Ai2 releases OlmoEarth v1.3 with global satellite embeddings
Ai2 says the explorer lets teams test mapping ideas in seconds before deciding whether a task warrants hours of fine-tuning.
• Ai2 released OlmoEarth v1.3 and precomputed annual satellite embeddings for the whole planet.
• Its browser-based explorer lets users make maps and compare land changes without signing in or training a model.
• Ai2 says v1.3 took 53% of v1.2’s computing time to train and produced 50% of its carbon emissions.
Read more
Ai2 says the explorer lets teams test mapping ideas in seconds before deciding whether a task warrants hours of fine-tuning.
• Ai2 released OlmoEarth v1.3 and precomputed annual satellite embeddings for the whole planet.
• Its browser-based explorer lets users make maps and compare land changes without signing in or training a model.
• Ai2 says v1.3 took 53% of v1.2’s computing time to train and produced 50% of its carbon emissions.
Read more
Google introduces Agent Compute Engine and expands agent governance
Google says the new service and governance tools will help organizations run isolated agents and apply policies across more environments.
• Google introduced Agent Compute Engine for AI agents, with a preview planned in the coming months.
• Its open-source Agent Substrate runtime is available to try on Google Kubernetes Engine today.
• Google says expanded agent governance will reach other clouds and managed devices, with previews planned soon.
Read more
Google says the new service and governance tools will help organizations run isolated agents and apply policies across more environments.
• Google introduced Agent Compute Engine for AI agents, with a preview planned in the coming months.
• Its open-source Agent Substrate runtime is available to try on Google Kubernetes Engine today.
• Google says expanded agent governance will reach other clouds and managed devices, with previews planned soon.
Read more
Nvidia launches Minerva to monitor AI factories
Nvidia says Minerva can help AI factory operators trace workload slowdowns to infrastructure problems and focus on fixes.
• Nvidia launched Minerva to help operators find problems across AI factory infrastructure and workloads.
• Nvidia says it uses AI models to detect anomalies, predict failures and group related alerts.
• IREN is using Minerva, and seven other partners plan to deploy it.
Read more
Nvidia says Minerva can help AI factory operators trace workload slowdowns to infrastructure problems and focus on fixes.
• Nvidia launched Minerva to help operators find problems across AI factory infrastructure and workloads.
• Nvidia says it uses AI models to detect anomalies, predict failures and group related alerts.
• IREN is using Minerva, and seven other partners plan to deploy it.
Read more
OpenAI seeks up to $30B at a $1.25T valuation
If the round values OpenAI above $1 trillion, it would set a private-company record and raise the valuation benchmark for its planned IPO.
• OpenAI is seeking up to $30 billion at a $1.25 trillion pre-money valuation, sources told CNBC.
• Coatue and UAE investors would lead the round, whose terms are still being negotiated.
• OpenAI filed confidential IPO paperwork in June after a March round valued it at $852 billion post-money.
Read more
If the round values OpenAI above $1 trillion, it would set a private-company record and raise the valuation benchmark for its planned IPO.
• OpenAI is seeking up to $30 billion at a $1.25 trillion pre-money valuation, sources told CNBC.
• Coatue and UAE investors would lead the round, whose terms are still being negotiated.
• OpenAI filed confidential IPO paperwork in June after a March round valued it at $852 billion post-money.
Read more
Anthropic says Claude verified tests across 123 open-source libraries
Under the stated RISC-V and Linux assumptions, the proofs cover every input smaller than 64 MB for the verified test entry points.
• Anthropic says Claude proved 239 test entry points across 123 open-source libraries cannot crash under specified conditions.
• The proofs cover inputs smaller than 64 MB on modeled RISC-V binaries running under Linux.
• Anthropic is publishing some proofs, and Theorem is offering the process to other open-source projects.
Read more
Under the stated RISC-V and Linux assumptions, the proofs cover every input smaller than 64 MB for the verified test entry points.
• Anthropic says Claude proved 239 test entry points across 123 open-source libraries cannot crash under specified conditions.
• The proofs cover inputs smaller than 64 MB on modeled RISC-V binaries running under Linux.
• Anthropic is publishing some proofs, and Theorem is offering the process to other open-source projects.
Read more
Meta plans up to $95 billion in stock sales for AI infrastructure
Meta's planned stock sales would fund its AI infrastructure buildout, while its new Class C shares would carry no voting rights.
• Meta plans up to $95 billion in stock sales for AI infrastructure, according to The Information.
• Meta says $22.5 billion would come from Class A stock and $12.5 billion from depositary shares.
• The remaining $60 billion would come from open-market sales after Class C shares start trading, likely no earlier than Q1 2027.
Read more
Meta's planned stock sales would fund its AI infrastructure buildout, while its new Class C shares would carry no voting rights.
• Meta plans up to $95 billion in stock sales for AI infrastructure, according to The Information.
• Meta says $22.5 billion would come from Class A stock and $12.5 billion from depositary shares.
• The remaining $60 billion would come from open-market sales after Class C shares start trading, likely no earlier than Q1 2027.
Read more
SpaceX secures roughly $40 billion to lease AI chips from Apollo
The lease gives SpaceX another funding source as it expands AI computing capacity and expects $35 billion to $40 billion in fourth-quarter capital spending.
• SpaceX secured a binding commitment for roughly $40 billion in AI chip and equipment financing.
• An Apollo-managed entity will buy the hardware and lease it to SpaceX for about five to five-and-a-quarter years.
• SpaceX expects the deal to close by year-end, with leases beginning as equipment arrives in the first half of 2027.
Read more
The lease gives SpaceX another funding source as it expands AI computing capacity and expects $35 billion to $40 billion in fourth-quarter capital spending.
• SpaceX secured a binding commitment for roughly $40 billion in AI chip and equipment financing.
• An Apollo-managed entity will buy the hardware and lease it to SpaceX for about five to five-and-a-quarter years.
• SpaceX expects the deal to close by year-end, with leases beginning as equipment arrives in the first half of 2027.
Read more
Cloudflare fixes cross-customer data exposure in GLM 5.3 on Workers AI
Cloudflare says the flaw put data from earlier customers' GLM 5.3 requests into 155 responses before the model was fixed.
• Cloudflare says 155 GLM 5.3 responses exposed data from other customers' requests between August 27 and September 9.
• All affected responses were to prompts longer than 681,280 tokens, according to Cloudflare.
• Cloudflare restored the model on September 9 after changing its configuration and patching SGLang.
Read more
Cloudflare says the flaw put data from earlier customers' GLM 5.3 requests into 155 responses before the model was fixed.
• Cloudflare says 155 GLM 5.3 responses exposed data from other customers' requests between August 27 and September 9.
• All affected responses were to prompts longer than 681,280 tokens, according to Cloudflare.
• Cloudflare restored the model on September 9 after changing its configuration and patching SGLang.
Read more
IBM adds agent identity and access controls to Vault
IBM says the controls help companies identify who owns an AI agent, limit what it can access and review its actions across business systems.
• IBM says Agent Identity and Vault 2.2 give AI agents tracked identities and limited credentials; both are generally available.
• Watsonx Orchestrate’s agent discovery and containment features are in tech preview.
• Orchestrate’s agent evaluation capabilities are generally available, with testing, monitoring and cost tracking.
Read more
IBM says the controls help companies identify who owns an AI agent, limit what it can access and review its actions across business systems.
• IBM says Agent Identity and Vault 2.2 give AI agents tracked identities and limited credentials; both are generally available.
• Watsonx Orchestrate’s agent discovery and containment features are in tech preview.
• Orchestrate’s agent evaluation capabilities are generally available, with testing, monitoring and cost tracking.
Read more
Nvidia plans $3 billion Meta investment as chip partnership expands
The reported deal adds to investor debate over Nvidia funding customers that build data centers with its chips.
• Nvidia plans to invest $3 billion in Meta’s planned equity offering, the Wall Street Journal reports.
• Meta plans to use more Nvidia hardware for AI agents and expand the companies’ partnership around Muse.
• Meta wants to raise up to $95 billion for AI infrastructure; AMD and Broadcom signaled nonbinding interest in shares.
Read more
The reported deal adds to investor debate over Nvidia funding customers that build data centers with its chips.
• Nvidia plans to invest $3 billion in Meta’s planned equity offering, the Wall Street Journal reports.
• Meta plans to use more Nvidia hardware for AI agents and expand the companies’ partnership around Muse.
• Meta wants to raise up to $95 billion for AI infrastructure; AMD and Broadcom signaled nonbinding interest in shares.
Read more
Google makes Gemini Robotics ER 2 available in public preview
In public preview, ER 2 gives developers a way to connect robot task planning and video understanding with controls and tools on Google Cloud.
• Google has made Gemini Robotics ER 2 available in public preview on the Gemini platform.
• The model helps robots plan tasks, interpret video and coordinate lower-level controls.
• Developers and enterprise teams can access it now through Model Garden.
Read more
In public preview, ER 2 gives developers a way to connect robot task planning and video understanding with controls and tools on Google Cloud.
• Google has made Gemini Robotics ER 2 available in public preview on the Gemini platform.
• The model helps robots plan tasks, interpret video and coordinate lower-level controls.
• Developers and enterprise teams can access it now through Model Garden.
Read more
Google DeepMind announces Genie 4 with real-time audio and interactive worlds
In research preview, Genie 4 lets testers steer simulations with text while the model generates visuals, sound and dialogue in real time.
• Google DeepMind is rolling out Genie 4 as a research preview for trusted testers.
• The real-time world model generates sound and interactive dialogue alongside its scenes.
• Later access will start with Google AI Ultra subscribers and early access Gemini API and Google AI Studio partners.
Read more
In research preview, Genie 4 lets testers steer simulations with text while the model generates visuals, sound and dialogue in real time.
• Google DeepMind is rolling out Genie 4 as a research preview for trusted testers.
• The real-time world model generates sound and interactive dialogue alongside its scenes.
• Later access will start with Google AI Ultra subscribers and early access Gemini API and Google AI Studio partners.
Read more
Google and Oracle plan to bring TPUs to Oracle Cloud in 2027
Starting in 2027, the planned deployment would give OCI customers a way to run AI training and inference on Google TPUs alongside GPUs.
• Google and Oracle plan to deploy TPU capacity on OCI starting in calendar year 2027.
• Google says OCI will be the first hyperscale cloud provider outside Google to offer TPUs.
• Google says OCI customers will be able to use TPUs for training and inference near their Oracle data.
Read more
Starting in 2027, the planned deployment would give OCI customers a way to run AI training and inference on Google TPUs alongside GPUs.
• Google and Oracle plan to deploy TPU capacity on OCI starting in calendar year 2027.
• Google says OCI will be the first hyperscale cloud provider outside Google to offer TPUs.
• Google says OCI customers will be able to use TPUs for training and inference near their Oracle data.
Read more
Nvidia and Oracle plan multi-gigawatt Vera Rubin clusters on Oracle Cloud
According to Nvidia, OCI customers can scale CPU and GPU resources independently for AI model workloads and tasks performed by agents.
• Nvidia and Oracle plan to expand OCI Superclusters with Vera Rubin NVL72 systems to multi-gigawatt scale.
• The companies say Vera Rubin NVL72 systems and Vera CPUs are generally available on OCI.
• OCI plans an eight-GPU HGX Rubin NVL8 configuration for x86-based servers.
Read more
According to Nvidia, OCI customers can scale CPU and GPU resources independently for AI model workloads and tasks performed by agents.
• Nvidia and Oracle plan to expand OCI Superclusters with Vera Rubin NVL72 systems to multi-gigawatt scale.
• The companies say Vera Rubin NVL72 systems and Vera CPUs are generally available on OCI.
• OCI plans an eight-GPU HGX Rubin NVL8 configuration for x86-based servers.
Read more
Anthropic says Claude found critical flaws in Chrome, Safari and operating systems
Anthropic says Claude combined flaws into attacks that could run code, expose personal data or crash servers with little or no user interaction.
• Anthropic says an internal Claude model built five exploit chains affecting Chrome, Safari, Android, Windows and Linux.
• Google patched 62 high or critical Chrome flaws reported by Anthropic as of Chrome 156.
• Anthropic says the Safari, Android, Windows and Linux issues have also been fixed.
Read more
Anthropic says Claude combined flaws into attacks that could run code, expose personal data or crash servers with little or no user interaction.
• Anthropic says an internal Claude model built five exploit chains affecting Chrome, Safari, Android, Windows and Linux.
• Google patched 62 high or critical Chrome flaws reported by Anthropic as of Chrome 156.
• Anthropic says the Safari, Android, Windows and Linux issues have also been fixed.
Read more