OpenAI unveils GPT-5.3-Codex-Spark as Google faces attacks

Key Takeaways

• OpenAI launched GPT-5.3-Codex-Spark on Cerebras chips, generating over 1,000 tokens per second, making it 15 times faster than its older version.
• Google's Gemini chatbot was targeted by over 100,000 "distillation attacks" by private companies and researchers attempting to clone its AI model.
• Google is offering voluntary exit packages to employees in its Global Business Organization who are not fully embracing AI and working with higher intensity.
• Microsoft is developing its own AI models and Maia 200 chips to reduce reliance on OpenAI and compete with Nvidia, aiming for greater control over its AI infrastructure.
• Investor Michael Burry remains skeptical about the current AI stock boom and Palantir Technologies, despite its stock reaching new highs.
• Investor concerns about AI automating truck driver jobs caused the Dow Jones Transport Index to drop 3.7%.
• The University of Nebraska launched a new systemwide AI institute focused on responsible and human-centered AI.
• New York lawmaker Alex Bores proposed an AI regulation plan for Congress, including requiring companies to report AI-related job losses and mandating independent safety testing for AI models.
• OpenAI co-founder Greg Brockman funds a super PAC that opposes Alex Bores' AI regulation plan.
• Google Gemini helped a couple resolve a marital dispute by providing impartial analysis and suggestions.

OpenAI launches super fast coding AI on Cerebras chips

OpenAI released its new coding model, GPT-5.3-Codex-Spark, which runs on Cerebras chips instead of Nvidia hardware. This model is incredibly fast, generating over 1,000 tokens per second, making it about 15 times quicker than its older version. OpenAI is making this research preview available to ChatGPT Pro subscribers through the Codex app and other tools. This move shows OpenAI's effort to rely less on Nvidia and explore new chip partnerships.

OpenAI unveils super fast AI coding model Spark

OpenAI launched a research preview of GPT-5.3 Codex-Spark, an AI coding model built for extreme speed. It runs on Cerebras Wafer-Scale Engine 3 hardware, a single large chip that helps it deliver over 1,000 tokens per second, making it 15 times faster than the flagship GPT-5.3 Codex. OpenAI also improved software to reduce delays, allowing for real-time steering of the model. While incredibly fast, Spark is a smaller model with less reasoning depth and is not recommended for sensitive security tasks. ChatGPT Pro users can access this new tool.

OpenAI unveils GPT-5.3-Codex-Spark for fast coding

OpenAI introduced GPT-5.3-Codex-Spark on February 12, 2026, as its first model built for real-time coding. This research preview, a smaller version of GPT-5.3-Codex, is the first result of OpenAI's partnership with Cerebras. It delivers over 1,000 tokens per second, making coding feel almost instant. ChatGPT Pro users can now try this model, which is designed for quick, interactive work and allows users to interrupt or redirect it in real time. OpenAI also made system-wide latency improvements that will benefit all its models.

Google's Gemini chatbot faces cloning attempts

Google reported that its AI chatbot, Gemini, has been targeted by "commercially motivated" attackers trying to clone it. These attackers used "distillation attacks," repeatedly asking Gemini questions, with one campaign sending over 100,000 prompts. Their goal is to uncover Gemini's internal logic and patterns to create or improve their own AI models. Google believes private companies and researchers are behind these global attacks, which the company views as intellectual property theft. John Hultquist from Google's Threat Intelligence Group warned that smaller companies using custom AI tools could also face similar threats.

Google reports 100,000+ attempts to clone Gemini AI

Google announced that its AI model, Gemini, was targeted by attackers who prompted it over 100,000 times to create a copycat. This process, called "distillation," involves feeding an AI model many prompts to learn its behavior and then training a new, smaller model from its outputs. Google views this "model extraction" as intellectual property theft, though the legal line is untested. Private companies and researchers worldwide are believed to be behind these efforts, seeking a competitive advantage. Other companies like OpenAI have also faced similar cloning attempts, highlighting a common challenge for publicly accessible AI models.

University of Nebraska opens new AI institute

The University of Nebraska has launched a new systemwide artificial intelligence institute. This institute aims to make Nebraska a national leader in developing responsible and human-centered AI. It will focus on research and education to shape the future of AI technology.

Michael Burry doubts AI stock boom and Palantir

Michael Burry, the investor known for predicting the 2008 financial crisis, remains doubtful about the current excitement around AI stocks. He is especially skeptical about Palantir Technologies, even though its stock has reached new highs. Burry's cautious view suggests he sees a gap between the market's high hopes and the actual value of these AI companies. He encourages investors to think carefully about the long-term future of the AI market.

Google offers buyouts to employees not focused on AI

Google is offering voluntary exit packages to some employees in its Global Business Organization. Chief Business Officer Philipp Schindler sent an email on Tuesday, February 10, stating that all staff must fully embrace AI and work with higher intensity. Employees who are not comfortable with the current pace or wish to leave can accept a severance package. This offer applies to teams like solutions, sales, and corporate development, but not customer-facing roles. Google uses buyouts as a voluntary way to reduce staff, avoiding the negative impact and legal issues of mass layoffs.

Microsoft develops its own AI models and chips

Microsoft is building its own AI models and chips to reduce its reliance on OpenAI. The company wants to ensure its "Copilot everywhere" strategy does not depend on a single supplier. While OpenAI remains a key partner, Microsoft aims to control its own AI infrastructure, including compute, security, and billing. Microsoft acquired an AI model from Inflection AI in August 2025 and is developing its Maia 200 chip to compete with Nvidia. The company also hosts models from other developers like xAI and Meta, even using Meta's Llama 2 for some Copilot tasks.

Accio 2026 launches AI search for B2B suppliers

Accio 2026 is a new AI-driven B2B search engine and commercial assistant designed for purchasing managers and entrepreneurs. It aims to solve common problems like outdated supplier directories, inefficient searches, and manual data management. Backed by Alibaba's vast data, Accio's goal is to automate the entire supplier workflow, not just the search. Its "Accio Agent" feature helps verify suppliers, centralize information, and streamline tedious tasks, making the process of finding and managing suppliers much more efficient.

Alex Bores reveals AI regulation plan for Congress

New York state lawmaker Alex Bores, a candidate for Congress, has unveiled his plan for regulating artificial intelligence. His agenda includes eight policy proposals, such as cutting red tape for renewable energy data centers and requiring companies to report AI-related job losses. Bores also calls for a national RAISE Act to mandate independent safety testing for AI models and create accountability for harm. Additionally, his plan addresses child safety standards for AI and proposes a national data privacy law. Bores faces opposition from some AI industry leaders, including a super PAC funded by OpenAI co-founder Greg Brockman, but also receives support from other AI workers.

Google Gemini helps couple resolve argument

Rachel Rood, a deputy editor and journalism professor, shared how Google Gemini helped her and her husband resolve a marital dispute. While initially wary of AI, she agreed to use Gemini after an argument about her husband's car breaking down late at night. Gemini analyzed their scenario, giving weight to both sides and suggesting Rachel could have shown more empathy. This AI's third-party input helped the couple understand each other better and move past their conflict. Ideastream Public Media is also hosting a community event to discuss AI's impact on Northeast Ohio.

AI fears cause trucking stocks to drop

Investor concerns about artificial intelligence are causing trucking and logistics stocks to fall. The Dow Jones Transport Index dropped 3.7% on Tuesday, with companies like XPO Logistics Inc. and Old Dominion Freight Line Inc. seeing significant declines. Investors worry that AI advancements, especially in autonomous driving, could automate many truck driver jobs. This potential disruption could lead to widespread job losses and retraining needs, impacting the entire economy. The market's reaction shows that the fear of AI's short-term impact on the transport sector is currently stronger than hopes for future growth.

Sources

• OpenAI sidesteps Nvidia with unusually fast coding model on plate-sized chips
• OpenAI Releases a Research Preview of GPT‑5.3-Codex-Spark: A 15x Faster AI Coding Model Delivering Over 1000 Tokens Per Second on Cerebras Hardware
• Introducing GPT-5.3-Codex-Spark
• Google says attackers used 100,000+ prompts to try to clone AI chatbot Gemini
• Attackers prompted Gemini over 100,000 times while trying to clone it, Google says
• University of Nebraska launches systemwide artificial intelligence institute
• Why Michael Burry remains skeptical around the AI trade, Palantir
• Google Offers Voluntary Exit Packages To Employees Not 'Embracing' AI
• Microsoft is building its own AI model
• Recensione Accio 2026: Il motore di ricerca B2B per fornitori
• The candidate at the center of the brewing midterm AI war unveils his agenda
• AI played marriage ref for my husband and me. What does it do for you?
• AI Panic Hits Trucking, Transport Stocks