Google Wins Approval for $10 Million Spirit Airlines Data Deal

Technology and AI stocks fell sharply, with Nvidia among the leaders in the losses, as a Closing Bell panel discussed the market selloff. Separately, NBC reported that former Biden security adviser urged urgent action against AI-enabled cyberattacks, and that Donald Trump wants AI chip sales to China to continue; the reports provided no further details.

A federal bankruptcy judge approved Google’s $10 million purchase of data from bankrupt Spirit Airlines for AI training. The records include business and operational material such as emails, Microsoft Teams messages, booking and transaction records, and some employee-related information. After objections from unions, privacy advocates and more than 120 lawmakers, the revised deal excludes sensitive employee information, requires independent de-identification and verification, and bars Google from knowingly trying to identify individuals. Google says it will remove identifying information and use the data to improve its AI models.

A 50-page House report warns that U.S. companies’ use of Chinese open-weight AI models could pose security risks. Reps. Andrew Garbarino and John Moolenaar say companies value the models for their low cost, customizability and performance, while warning that Chinese developers used proxy networks and fake accounts to copy U.S. models. Recommendations include a U.S. open-weight AI initiative, tracking model origins, and restricting Chinese models in federal contracts and sensitive infrastructure.

Two reports focus on AI governance and security. A report on the SEC says the agency has not issued AI-specific capital-markets rules, leaving questions about accountability, misleading information and dependence on a small number of developers. In cybersecurity, an analysis argues that AI agents need organizational context—such as systems, past investigations and risk limits—not simply more data. It recommends safeguards, human approval for higher-risk actions, ongoing testing and efficient models. Former Secret Service investigator Jason Brown also warned that voice cloning can enable impersonation scams, advising people to verify suspicious requests independently and consider a family safe word.

Research on time-series forecasting introduces Backtest-Gated Augmentation Averaging (BGAA), which tests different data views against a series’ own history and averages those that beat the original. It reduced forecasting error across the tested benchmarks; on BOOM, a 313-million-parameter model using BGAA outperformed 1-billion- and 2.5-billion-parameter models in the same family. Selecting the best view separately for each forecast window remains an open challenge.

Organizations are also setting out practical approaches to AI. Louisville Public Media says it uses or explores AI for behind-the-scenes tasks such as transcription, research and data analysis, with people reviewing AI-assisted work. It says journalists will continue reporting and writing, will not use AI to copy voices or alter real media, and retain editorial independence from contributions including Google News Initiative and OpenAI Foundation’s People-First AI Fund. In South Korea, SelectStar and the Korea Publication Culture Association plan a copyright-protected publishing-data system, with rights-holder consent, contracts, access controls and encryption; SelectStar will provide document-processing technology, including for closed or on-site systems.

At 16, Henrijs Dandzbergs founded Polemos, a startup developing AI-guided missiles. The company says its system would analyze a battlefield and select targets while remaining under remote human control, with the stated aim of reducing unintended harm. Dandzbergs has faced skepticism and technical challenges, while the startup has attracted investor and military attention.

Key Takeaways

  • Nvidia led losses as technology and AI stocks suffered a sharp market selloff.
  • A bankruptcy judge approved Google’s $10 million Spirit Airlines data purchase for AI training, with revised privacy safeguards and sensitive employee information excluded.
  • A House report warns of security risks from U.S. companies using Chinese open-weight AI models and recommends tracking model origins and limiting their use in sensitive federal settings.
  • A report says the SEC has no AI-specific capital-markets rules and highlights unresolved questions about accountability and other risks.
  • AI security agents need organizational context, risk-based limits and human oversight; Jason Brown warns that AI voice cloning is also enabling impersonation scams.
  • The BGAA forecasting method reduced error across tested benchmarks; a 313-million-parameter model using it beat larger models in its family on BOOM.
  • Louisville Public Media uses AI for support tasks while keeping editorial work and review in human hands; it says technology funding does not influence its editorial choices.
  • SelectStar and the Korea Publication Culture Association plan a publishing-data system requiring rights-holder consent and using access controls and encryption.
  • Sixteen-year-old Henrijs Dandzbergs founded Polemos to develop remotely controlled, AI-guided missiles.
  • NBC reports that a former Biden adviser called for urgent action on AI cyberattacks and that Donald Trump wants AI chip sales to China to continue; the reports offer no further detail.

Nvidia and AI stocks lead sharp market selloff

A Closing Bell panel discussed a sharp selloff that hit technology and artificial intelligence stocks. Nvidia was among the names leading the losses. Courtney Garcia of Payne Capital Management and Malcolm Ethridge of Capital Area Planning Group joined the program to react to the day’s market moves.

Former Biden adviser urges action against AI cyberattacks

A former security adviser to President Joe Biden called for urgent action to stay ahead of cyberattacks involving artificial intelligence. The item was published by NBC on November 20, 2026. The available article text does not provide further details about the adviser’s comments.

Hugo Lowell says Trump wants AI chip sales to China to continue

NBC’s Hugo Lowell reports that Donald Trump wants AI chip sales to China to continue. The item is dated November 19, 2026. The available article text does not provide further details about the report.

Google wins approval to buy Spirit Airlines data for AI training

A federal bankruptcy judge in New York approved Google’s $10 million deal to buy data from bankrupt Spirit Airlines. The information includes internal emails, spreadsheets, booking and frequent flyer details, transaction records, and employee data. Google says it will use the data to improve its artificial intelligence models and will remove identifying information. Spirit stopped operating after the Trump administration did not reach a bailout agreement, and a flight attendants’ union objected to the data sale in August.

Google’s $10 million Spirit Airlines data deal gets bankruptcy approval

A U.S. bankruptcy judge approved Google’s $10 million purchase of former Spirit Airlines business data on November 17, 2026. The records include emails, Microsoft Teams messages, employee and training information, and operational and business data, though reports give different estimates of the dataset’s size. Labor unions, privacy advocates, and more than 120 U.S. lawmakers raised concerns that former employees could be identified from the records. The revised deal excludes sensitive employee information and adds independent de-identification and verification safeguards, while barring Google from knowingly trying to identify people in the transferred data.

House report warns against US companies using Chinese AI models

A 50-page congressional report by Reps. Andrew Garbarino and John Moolenaar warns that US companies’ use of China-developed open-weight AI models could create security risks. The lawmakers received briefings from DoorDash, Airbnb and Anysphere about how they use these models. The report says companies choose them because they are inexpensive, easy to customize and effective, but warns that Chinese developers used proxy networks and fake accounts to copy the abilities of US AI models. Its 12 recommendations include creating a US open-weight AI initiative, tracking where models come from, and blocking Chinese models from federal contracts and sensitive infrastructure.

A history based verifier boosts time series forecasting

Researchers propose Backtest-Gated Augmentation Averaging, or BGAA, a test-time computing method for time series foundation models. It uses rolling backtests on a series’s own history to score different views of the data, removes views that perform worse than the original, and averages the remaining forecasts. BGAA lowered MASE across all tested model and benchmark combinations, including by 0.5 to 1.0% on GIFT-Eval, 1.3 to 5.0% on fev-bench, and 1.7 to 5.9% on BOOM. On BOOM, a 313 million parameter model using the method outperformed 1 billion and 2.5 billion parameter models in its family. The findings also show that choosing the best view for each test window could improve results much more, leaving window-level view selection as an open challenge.

AI security agents need context, not just scale

Security teams face huge volumes of alerts, and AI agents can help investigate them, but handling more data does not automatically improve threat detection. The article argues that agents need knowledge of each organization’s systems, past investigations, procedures and risk limits to make useful decisions. Companies can provide this context through methods such as model fine-tuning, retrieval, memory, workflow tools and policy controls, and should test agents as their systems change. Agents also need clear limits on what they can do alone, when people must approve actions, and how findings move between teams. To keep costs manageable, organizations should use efficient models and match an agent’s independence to the risk of its task.

SEC report highlights unresolved questions about AI in finance

A new report says the Securities and Exchange Commission has not issued rules specifically for artificial intelligence in capital markets, even as financial firms use AI in trading, compliance and client services. It points to unanswered policy concerns, including who is accountable when AI systems fail. The report also raises concerns about the risks of relying on a small number of AI developers and about potentially misleading information. The supplied article excerpt does not provide further details about the report or its recommendations.

Louisville Public Media explains its approach to AI

Louisville Public Media says AI is part of its three-year plan for digital transformation, with the goal of saving staff time for higher-value work. The organization is exploring and using AI behind the scenes for tasks such as transcription, research and data analysis, and people review anything AI helps produce. LPM says its journalists will continue to report and write stories, its staff will choose music, and it will not use AI to copy voices or alter real audio, photos or video. It also says contributions from technology companies, including the Google News Initiative and the OpenAI Foundation’s People-First AI Fund, do not influence its editorial choices. LPM says people will continue to make and lead its content, guided by its professional standards and ethics.

SelectStar and publishing group plan copyright protected AI data system

SelectStar signed a memorandum of understanding with the Korea Publication Culture Association to build a system for supplying publishing data to train AI. They plan to prepare and check the data, set up a supply system, encourage data transactions, and take part in government and public sector AI data projects. Use of the data will require consent from copyright holders and publishing rights holders, with supply terms set in separate contracts. Security plans include access controls and encryption. SelectStar will contribute document processing technology that can prepare complex documents for AI training, including in closed networks or on-site systems without sending files outside.

At 16, Henrijs Dandzbergs starts AI missile company Polemos

Henrijs Dandzbergs, a 16-year-old high school dropout, founded Polemos, a startup developing missiles that use artificial intelligence to find and strike enemy forces. He began exploring AI and its use in warfare at age 14, then developed a missile prototype and secured investor funding. Polemos says its missile will analyze the battlefield, select targets, and be controlled remotely by a human operator, with the aim of reducing unintended harm. Dandzbergs has faced doubts about AI in warfare and technical challenges, but says he hopes to bring the missile to market. The article says the startup has drawn attention from investors and the military.

Former Secret Service investigator warns of AI voice and message scams

Former Secret Service investigator Jason Brown warned that AI voice cloning and impersonation scams make it harder to tell real messages from fake ones. He said recent technology outages involving Microsoft, the Federal Aviation Administration, and financial institutions have left people anxious, though some incidents involved infrastructure problems and others remained under investigation. Brown cautioned that scammers may imitate a relative using a short voice recording or claim a package needs extra postage, then demand payment in gift cards or cryptocurrency. He advised people to stop, think, and independently verify suspicious messages, and suggested families use a safe word to confirm identity during unexpected calls. He warned that these scams can cause major financial losses, especially as the holiday season approaches.

Sources

NOTE:

This news brief was generated using AI technology (including, but not limited to, Google Gemini API, Llama, Grok, and Mistral) from aggregated news articles, with minimal to no human editing/review. It is provided for informational purposes only and may contain inaccuracies or biases. This is not financial, investment, or professional advice. If you have any questions or concerns, please verify all information with the linked original articles in the Sources section below.

AI market selloff Nvidia AI training data Google Spirit Airlines data privacy Chinese open-weight AI AI security AI governance SEC regulation AI agents human oversight voice cloning scams time-series forecasting BGAA AI in journalism copyright-protected data AI-guided missiles AI cyberattacks AI chip exports to China

Comments

Loading...