OpenAI Snowflake and Google lead AI news

Key Takeaways

• OpenAI's GPT-6 Cyber scored 48.5% on ExploitGym, beating GPT-5.6 Cyber's 32.1%, and found 81 unknown Firefox vulnerabilities and 948 Linux kernel exploits.
• OpenAI classified GPT-6 Cyber as Critical under its Preparedness Framework and restricted access to approved Daybreak Red customers.
• GPT-6 Cyber avoided out-of-scope exploits 99.93% of the time without safety prompts and 100% with them.
• OpenAI partnered with Hacktron, discovered through its Security Bug Bounty program, to test vulnerabilities in its own systems and third-party integrations.
• Mozilla has released fixes for the Firefox vulnerabilities identified by GPT-6 Cyber during testing.
• Snowflake launched CoCo, an AI tool integrated into Excel and Google Sheets that respects governed data policies including role and row-access restrictions.
• Nvidia stock rebounded over 5% after a 9.3% drop linked to Democratic midterm gains and fears of tighter AI regulation.
• The Nasdaq lost about $1.2 trillion in value on Thursday as investors feared stricter AI and data center rules under a Democratic-controlled Congress.
• CERT-In released AI security auditing guidelines with a tiered risk framework covering nine core domains including governance, data privacy, and human oversight.
• Schellman CEO Avani Desai flagged a gap in the White House AI oversight accord, noting it lacks definitions for who conducts assessments and how independence is ensured.

OpenAI launches GPT-6 Cyber AI model to find software flaws

OpenAI released GPT-6 Cyber, an AI model trained to find and fix software vulnerabilities. In testing, it discovered 81 unknown high-severity Firefox flaws and helped develop exploits for 948 Linux kernel vulnerabilities. The model scored 48.5 percent on ExploitGym, beating GPT-5.6 Cyber's 32.1 percent. OpenAI classified it as Critical under its Preparedness Framework and is restricting access to approved Daybreak Red customers for now.

OpenAI partners with Hacktron to hunt system vulnerabilities

OpenAI announced a partnership with Hacktron to find vulnerabilities in its own systems. The teams will test high-priority areas, especially how OpenAI systems interact with third-party services. Hacktron builds AI agents for code review and penetration testing and was discovered through OpenAI's Security Bug Bounty program. OpenAI is also expanding Hacktron's access to its frontier models.

OpenAI launches GPT-6 Cyber after finding Firefox and Linux flaws

OpenAI unveiled GPT-6 Cyber, a cybersecurity-focused model that found 81 unknown Firefox vulnerabilities and 948 Linux privilege escalation exploits. It scored 48.5 percent on ExploitGym, a benchmark for turning vulnerabilities into working exploits, compared to 32.1 percent for GPT-5.6 Cyber. OpenAI added guardrails to keep the model within authorized scope and reported it avoided out-of-scope exploits 99.93 percent of the time without safety prompts and 100 percent with them.

OpenAI's GPT-6 Cyber AI finds 81 Firefox flaws and 948 Linux bugs

OpenAI revealed GPT-6 Cyber, which found 81 previously unknown high-severity Firefox vulnerabilities and developed working exploits for 948 Linux kernel flaws. In benchmark tests, it exploited 48.5 percent of vulnerabilities on ExploitGym versus 32.1 percent for GPT-5.6 Cyber. OpenAI classified it as Critical under its Preparedness Framework and limited access to select Daybreak Red customers. The company warned defenders have a limited window to use powerful AI tools for security before attackers gain similar abilities.

GPT-6 Cyber advances aligned cybersecurity agent research

OpenAI introduced GPT-6 Cyber, a model designed to find exploitable vulnerabilities while staying within researcher-set boundaries. It achieved a 51 percent relative improvement on ExploitGym over GPT-5.6 Cyber and avoided out-of-scope exploits 99.93 percent of the time without deployment prompts and 100 percent with them. The model helped uncover 81 unknown Firefox flaws, and Mozilla has released fixes. GPT-6 Cyber also outperformed GPT-6 Astra on most cyber tasks and is classified as Critical under OpenAI's Preparedness Framework.

CERT-In releases tiered AI security auditing framework

CERT-In released AI Security Auditing Guidelines on November 4 that use a risk-based framework for auditing AI systems in India. The rules cover AI systems that process Indian data or serve Indian users and span nine core audit domains including governance, data privacy, bias, security and human oversight. AI systems are sorted into four risk tiers: Low, Moderate, High and Critical. Audit requirements vary by tier, with Moderate risk systems needing annual independent audits and High risk systems requiring more frequent assessments.

Schellman CEO highlights gap in White House AI oversight deal

Schellman CEO Avani Desai pointed out a major gap in the White House AI accord on independent assessments. The agreement calls for independent oversight but does not define who can conduct assessments, what independence means or how conflicts of interest should be handled. Desai stressed that independence must be demonstrated through clear scope, qualified experts, disclosure of commercial relationships and independent quality review. She also warned that oversight fails when assessors lack the freedom to report serious findings without pressure to soften them.

Snowflake brings governed AI into Excel and Google Sheets

Snowflake launched CoCo, an AI tool that works directly inside Excel and Google Sheets using governed Snowflake data. CoCo runs queries in Snowflake warehouses and streams results to spreadsheets while respecting role, masking and row-access policies. It removes ODBC and third-party connectors and keeps query text and formula lineage visible in the workbook. Early customer feedback highlighted improved correctness on complex data and faster report building compared to exporting data to other tools.

How AI-driven consumer choices may reshape brands

This article explores what happens to brands as AI search and AI agents become the main way consumers decide which products to buy. The author believes AI-assisted purchasing will become the dominant decision-making path for consumers. The piece raises questions about how brand visibility and marketing strategies will shift when AI agents, rather than humans, choose products. The article signals a major shift in how brands will need to compete for attention.

BITZER adds AI agent HERMANN to selection software

BITZER released version 7.2.0 of its selection software with an expanded SmarTube module for shell-and-tube condensers and evaporators. The update allows direct selection of water-cooled condensing units from the K series, covering 178 units in six variants. A new AI agent named HERMANN answers questions about BITZER products, refrigerants and applications and directs users to relevant technical documentation. The software is available as a web and Windows application and has supported refrigeration and HVAC professionals for over 35 years.

Nvidia and AI chip stocks rebound after big drop

Nvidia stock rose more than 5% on Friday morning after a major sell-off the day before. The chip company had crashed 9.3% on Thursday during a tech stock rout. Other AI-related chip stocks also recovered in early trading. The rebound came as investors regrouped after the sharp losses.

AI clones of GTA VI could push faster release

A former Rockstar Games tech director said AI clones of Grand Theft Auto VI could pressure the company to release the game sooner. The comment came when discussing the game's release date. AI clones are still a new idea and it is unclear when the technology will exist. Rockstar Games has not given an official release date. Fans must keep waiting for more details.

Klipboard launches AI sales assistant for car parts

Klipboard introduced an AI sales assistant built into its AutoCat software for automotive aftermarket businesses. The tool helps sales teams identify the right parts faster by asking questions in natural language. It guides users through complex enquiries and reduces avoidable returns. The assistant supports existing staff expertise rather than replacing it. Wayne Fraser, chief information officer at Klipboard, said the tool helps solve real operational challenges.

Global AI stocks fall after Democrats win midterms

Global stock markets in Europe and Asia dropped on Friday after big Democratic gains in the U.S. midterm elections. Investors feared that a Democratic-controlled Congress could bring tighter regulations for the AI industry. The Nasdaq fell 5.52% on Thursday, losing about $1.2 trillion in value. Silicon Valley executives including Elon Musk and Greg Brockman spent over half a billion dollars backing Republican candidates. Wall Street is now watching how the new Congress will approach AI and data center rules.

Tech stocks fall as AI trade worries grow

Technology stocks stayed under selling pressure on Friday as investors reacted to Democrats' strong showing in the U.S. midterm elections. The Bursa Malaysia Technology Index fell more than 6% at the opening bell. The broader FBM KLCI market also dropped more than 1.25%. Analyst Kenneth Leong said the sell-off was largely sentiment-driven and reflected reassessed valuations. He said longer-term growth drivers from AI adoption and semiconductor demand remain intact.

Sources

• OpenAI’s GPT-6 Cyber found 81 Firefox flaws earlier AI testing missed
• Partnering with Hacktron to find vulnerabilities in OpenAI’s systems
• OpenAI Launches GPT-6 Cyber After AI Helps Find 81 Firefox Flaws, 948 Linux Exploits
• OpenAI's new GPT-6 Cyber AI finds 81 Firefox flaws, 948 Linux bugs
• GPT‑6 Cyber: Advancing the frontier of aligned cybersecurity agents
• CERT-In’s AI Security Auditing Guidelines set out tiered framework
• Schellman CEO: the giant loophole in the White House AI deal
• Snowflake puts governed AI inside Excel and Sheets
• What happens to brands when AI helps consumers choose?
• BITZER Expands Selection Software 7.2.0 with AI Agent HERMANN
• Nvidia Stock, Other AI Names Rebound After Sell-Off
• Could potential AI clones of GTA VI force a quicker official port?
• Klipboard introduces AI sales assistant
• Global AI Stocks React to Democrats' U.S. Midterms Win
• Tech stocks tumble as AI trade concerns deepen