SendTech Times
العربية
AINews|July 15, 2026, 09:22 AM
CAPACITY TEST:

Grok Build Uploads Stop After Cereblab Flags Whole Repos

BySendTech Times AI & Enterprise DeskNewsroom-edited, source-reviewed coverage|Source: The Register
Newsroom brief

The Register reported that Grok Build stopped whole-repository uploads after Cereblab found the AI coding tool sending repo bundles and Git history to Google Cloud Storage. SpaceXAI and Elon Musk said user data would be deleted, but the report did not verify deletion or name an audit record.

Verified against source materialEdited by SendTech Times AI & Enterprise Desk
Grok Build Uploads Stop After Cereblab Flags Whole Repos

Grok Build stopped sending whole code repositories to SpaceXAI servers after a Cereblab report found that the AI coding tool uploaded repo bundles to Google Cloud Storage, The Register reported.

Elon Musk separately promised deletion of previously uploaded user data; the outlet left that claim unverified.

The case puts an AI coding assistant privacy dispute around source-code retention rather than model capability.

In Cereblab's test, the upload behaviour continued even when a prompt told the command-line tool to reply only with “OK” and not open files.

Grok Build Stopped Whole-Repo Uploads After A Server Flag

The repository uploads stopped after developers set the server flag disablecodebaseupload to “true,” according to The Register.

That server-side change was separate from SpaceXAI’s public advice to use the /privacy command.

Cereblab described /privacy as a per-session retention toggle, not the switch that stopped whole-repository uploads.

The researcher wrote that the right default should be off.

Cereblab Said Git Histories And Deleted Secrets Were Uploaded

Cereblab’s Sunday report described Grok Build transmitting unredacted file contents to a Google Cloud Storage bucket used by SpaceXAI.

The investigation also found that the tool packaged full repositories as Git bundles rather than sending only the files needed to answer a prompt.

The test repository included Git history with secrets deleted months earlier, according to The Register.

Other users reported similar results, including one case involving a user directory that contained SSH keys and password-manager databases.

SpaceXAI Said Zero Data Retention Customers Were Respected

SpaceXAI said through a public X statement that teams using zero data retention had no trace or code data retained.

It also said API key use of Grok Build respected zero data retention settings.

For users without zero data retention enabled, SpaceXAI pointed to the /privacy command inside the CLI as a way to disable data retention and delete previously synced data.

Musk later said all user data uploaded before the code change would be “completely and utterly deleted.”

The public record still lacks verification that SpaceXAI completed the promised deletion, along with an audit log, affected-user count, repository count, deletion timetable or third-party validation.

Share this article
inXf

Related articles

More
National Grid Backs Joulent With $1.75 Billion As AI Power Queues Stretch
Cloud & Data Centers

National Grid Backs Joulent With $1.75 Billion As AI Power Queues Stretch

National Grid Ventures plans a $1.75 billion minority investment in Joulent, a power developer built around dedicated generation for AI data centres. The deal supports Project Kilby in Texas, but delivery and interconnection risks remain outside the disclosed timetable.

Microsoft Human Rights Review Puts Cloud and AI Contracts Under Pre-Contract Scrutiny
AI

Microsoft Human Rights Review Puts Cloud and AI Contracts Under Pre-Contract Scrutiny

Microsoft said it will strengthen human rights controls after reviewing how the Israeli military used its technology during the Gaza war. The company said it had disabled specified cloud storage and AI service subscriptions for the Israeli Ministry of Defence in September last year. The practical question is whether stronger pre-contract reviews change how sensitive cloud and AI engagements are approved before deployment.

Micron And Anthropic Put AI Memory Supply Inside Claude’s Compute Plan
Chips & Semiconductors

Micron And Anthropic Put AI Memory Supply Inside Claude’s Compute Plan

Micron and Anthropic signed a strategic agreement covering AI memory and storage architecture, supply, Claude adoption inside Micron and a Series H investment.

Arm and Supermicro Put Agentic AI Servers to a CPU Test
Chips & Semiconductors

Arm and Supermicro Put Agentic AI Servers to a CPU Test

Supermicro has introduced new server platforms built around Arm’s AGI CPU for inference-heavy and agentic AI workloads across cloud, enterprise and edge deployments. Arm says the AGI CPU includes up to 136 Arm Neoverse V3 cores, 12 DDR5 memory channels running at up to 8800 MT/s and PCIe Gen6 connectivity within a 300W power envelope. The key test is whether operators can use these CPU-heavy designs to add inference capacity without creating new pressure on power and cooling.

New York Blocks New 50 MW Data Centre Permits For Up To A Year
Cloud & Data Centers

New York Blocks New 50 MW Data Centre Permits For Up To A Year

The Verge reported that New York Governor Kathy Hochul signed an executive order blocking new environmental permits for hyperscale data centres above 50 megawatts for up to a year. The report said a stricter 20 megawatt legislative moratorium still awaits her decision.

AI Compute Scarcity Is Redrawing The Infrastructure Map
AI

AI Compute Scarcity Is Redrawing The Infrastructure Map

AI infrastructure projects in India, Africa, Brazil and the UAE show how power, chip access, data location and inference demand are pushing compute beyond the traditional U.S. hyperscale cloud map.

Keep Reading

More Stories

Latest
Sheetz Moves 11,000 VMs Off VMware After Broadcom Licensing ShiftCloud & Data CentersJul 16, 2026Sheetz Moves 11,000 VMs Off VMware After Broadcom Licensing ShiftArs Technica reported that Sheetz is moving around 11,000 VMs across 838 locations from VMware vSphere to StorMagic SvHCI. The convenience-store chain has migrated more than 600 stores, while contract values, software licensing costs and final support terms remain outside the report.Revolut Adds VARA Crypto Approval To UAE Payments LicencesFintech & Digital PaymentsJul 16, 2026Revolut Adds VARA Crypto Approval To UAE Payments LicencesPYMNTS reported that Revolut received in-principle VARA approval for a UAE VASP licence, weeks after Central Bank of the UAE payment licences. PYMNTS said final approval is still required, with the UAE launch date, crypto pricing and first customer numbers outside the public record.US Export-Control Shift Opens UAE AI Chip Access, DCD ReportsChips & SemiconductorsJul 16, 2026US Export-Control Shift Opens UAE AI Chip Access, DCD ReportsData Center Dynamics reported that the US government has moved the UAE into a lower-restriction export-control category, giving UAE companies broader access to advanced AI chips from Nvidia and AMD. The account names G42 and hyperscaler data-centre projects, but it does not list chip volumes, approved licences or specific UAE projects.Cadence Adds AuraStack AI Agent For PCB And Advanced Packaging DesignChips & SemiconductorsJul 16, 2026Cadence Adds AuraStack AI Agent For PCB And Advanced Packaging DesignThe Register reported that Cadence Design Systems introduced AuraStack, an agentic AI system for PCB and advanced packaging workflows. Cadence cited a 15x productivity claim and named Nvidia among customers, but The Register did not include pricing, availability dates, full customer names or independent benchmark results.OpenAI Backs State AI Safety Baseline As Federal Cyber Tests NearCapital & PolicyJul 16, 2026OpenAI Backs State AI Safety Baseline As Federal Cyber Tests NearOpenAI said California, New York and Illinois have advanced frontier AI safety legislation with shared disclosure, incident-reporting and audit elements, while a federal cyber-testing framework is still targeted for early August.Intel Ships Panther Lake Chips Built With ASML High NA EUVChips & SemiconductorsJul 16, 2026Intel Ships Panther Lake Chips Built With ASML High NA EUVIntel has shipped selected Panther Lake processors made on Intel 18A with ASML High NA EUV lithography, Interesting Engineering reported, giving Intel production data while customer names, shipment volumes and wider node timing remain undisclosed.Thinking Machines Releases Inkling Open-Weights Model With 975 Billion ParametersAIJul 16, 2026Thinking Machines Releases Inkling Open-Weights Model With 975 Billion ParametersSiliconANGLE reported that Thinking Machines Lab has released Inkling, its first foundation model, with full open weights and fine-tuning through Tinker. The account cited 975 billion total parameters, about 41 billion active parameters per average prompt and training on about 45 trillion tokens, while leaving customer deployments and independent benchmark validation undisclosed.OpenAI Keeps GPT-Red Attack Model Private After Prompt-Injection TestsAIJul 16, 2026OpenAI Keeps GPT-Red Attack Model Private After Prompt-Injection TestsThe Next Web reported that OpenAI has built GPT-Red, an internal automated red-team model for prompt-injection attacks, but is keeping the attacker private. The report cited attack success rates above 90% against an older GPT-5 and below 23% against GPT-5.6, while noting that human testers still catch cases GPT-Red misses.Nokia And NVIDIA Push AI-RAN Roadmap Toward 2027 Field TestsChips & SemiconductorsJul 16, 2026Nokia And NVIDIA Push AI-RAN Roadmap Toward 2027 Field TestsNokia announced an AI-native RAN platform with NVIDIA Aerial integration and claimed demonstrated spectral-efficiency gains of more than 20%, while its larger 50% and 100% capacity targets remain tied to 2027 and 2028 roadmap milestones.OpenAI Adds Usage Analytics And Spend Controls For ChatGPT WorkAIJul 15, 2026OpenAI Adds Usage Analytics And Spend Controls For ChatGPT WorkOpenAI said GPT-5.6 uses 54% fewer output tokens and 57% less time per task in a named coding-agent index, while its enterprise guidance tells ChatGPT Work admins to manage AI spend by accepted outcomes, usage analytics and governance controls rather than token price alone.IMEC Outlines CMOS 2.0 Path As AI Compute Demand RisesAIJul 15, 2026IMEC Outlines CMOS 2.0 Path As AI Compute Demand RisesCommonWealth Magazine English reported that IMEC's new CEO Patrick Vandenameele outlined semiconductor roadmap work tied to AI inference demand, CMOS 2.0 stacking, memory placement and optical interconnects. CommonWealth reported that Vandenameele estimated a 150-fold workload increase as AI shifts from training to inference, while TSMC's Kevin Zhang pointed to nanosheet, CFET and 3D stacking work.Allstate Says Broadcom Ordered Four Audits After VMware ExitCapital & PolicyJul 15, 2026Allstate Says Broadcom Ordered Four Audits After VMware ExitArs Technica reported that a June filing accused Broadcom of starting four audits after Allstate chose not to renew VMware and CA contracts. VMware and CA have separate cases against the insurer, while the filings leave its current virtualisation stack and production impact undisclosed.