OpenAI and Anthropic dropped new, cheaper AI models within hours of each other. OpenAI introduced GPT-6 Sol and GPT-6 Luna, while Anthropic unveiled Claude Opus 5.5, a model built for long, complex tasks like codebase migrations and financial analysis. Opus 5.5 costs 40% less to run than its predecessor and always uses a reasoning process. Both companies also urged a measured pace for frontier AI development.
Anthropic added new safeguards to Opus 5.5 against distillation attacks, prompt-injection attacks, and other security risks. The move signals that cost reduction and safety are now advancing side by side, even as the industry pushes models toward more autonomous, real-world use cases.
Meta ran into a problem with its Muse AI assistant when merchants started hanging up on calls after learning the caller was AI. The company tested using human callers internally but paused the effort after employees flagged data leak and security concerns. Meta said it will only roll out the calling feature with proper disclosures in place.
In the cybersecurity space, Lookout launched Social Engineering Protection on September 23, 2026, targeting AI-powered attacks across mobile channels. The tool uses intent-driven AI, phone verification, and voice analysis to stop smishing, vishing, and voice cloning in real time. It works across SMS, MMS, RCS, WhatsApp, and voice calls on both iOS and Android.
SPS Commerce made its AI supply chain layer, MAX, generally available to tens of thousands of fulfillment customers. The platform covers fulfillment, onboarding, analytics, and supply chain visibility, and is being embedded directly into operations to cut down on manual setup. The company serves more than 300,000 trading relationships.
Google updated its Intelligent Endpoints strategy with agentic task automation and a Gemini Skills library in Chrome. The update includes clipboard DLP protection that blocks unauthorized copy actions at the trigger level. Legacy applications can now be streamed into Chrome tabs via Cameyo by Google, and IT teams gain GenAI usage reporting along with real-time visibility into web-based AI usage across mobile devices.
Big Tech AI spending continues to climb despite cash flow strain. A September 22, 2026 analysis by Nice Credit Rating found that Alphabet, Amazon, Meta, Microsoft, and Oracle face weakened cash flows from aggressive AI capital expenditures. The five companies held approximately 2.7 trillion dollars in confirmed contracts, though Oracle carries the heaviest burden with an order backlog equal to 9.5 years of revenue.
Voters are increasingly turning to AI for political information. OpenAI's ChatGPT and Google's AI tools have become frequent sources for election-related queries, and researchers say AI is playing a larger role in shaping the information environment than in past elections. Experts urge voters to stay skeptical about AI-generated outputs while acknowledging their growing influence.
Okta is betting on identity management as a way to control AI agents in enterprise settings, believing that verifying who and what accesses systems is key to safe deployment. A 2026 report from ESET found that 49% of UK small businesses experienced a cyberattack in the past year, with AI-powered malware and phishing as top concerns. Over half of those businesses plan to increase cybersecurity budgets in the next 12 months.
Key Takeaways
- <ul><li>OpenAI launched GPT-6 Sol and GPT-6 Luna while Anthropic unveiled Claude Opus 5.5, both targeting everyday business tasks at lower cost.</li><li>Claude Opus 5.5 costs 40% less to operate than Opus 5.1 and includes new safeguards against distillation and prompt-injection attacks.</li><li>Meta paused internal testing of human callers for its Muse AI assistant after employees raised data leak and security concerns.</li><li>Lookout launched Social Engineering Protection on September 23, 2026, to combat AI-powered smishing, vishing, and voice cloning on mobile devices.</li><li>SPS Commerce made its MAX AI supply chain layer generally available to tens of thousands of fulfillment customers.</li><li>Google updated Intelligent Endpoints with agentic task automation, a Gemini Skills library in Chrome, and clipboard DLP protection.</li><li>Alphabet, Amazon, Meta, Microsoft, and Oracle face weakened cash flows from aggressive AI spending, holding approximately 2.7 trillion dollars in confirmed contracts.</li><li>Voters increasingly rely on OpenAI's ChatGPT and Google's AI tools for election-related information, reshaping the information environment.</li><li>Okta is focusing on identity management to control AI agents, while 49% of UK small businesses experienced a cyberattack in the past year.</li><li>Over half of UK small businesses plan to increase cybersecurity budgets in the next 12 months in response to AI-powered threats.</li></ul>
OpenAI and Anthropic launch cheaper AI models after urging caution
OpenAI and Anthropic launched lower-cost AI models within hours of each other. OpenAI introduced GPT-6 Sol and GPT-6 Luna, while Anthropic unveiled Claude Opus 5.5. The new models target everyday business tasks and cost significantly less to operate, even as both companies urge a measured pace for frontier AI development.
Claude Opus 5.5 cuts costs and adds safeguards for autonomous AI
Anthropic released Claude Opus 5.5, designed for long and complex tasks like codebase migrations and financial analysis. The model costs 40% less to operate than Opus 5.1 and always uses a reasoning process. It includes new safeguards against distillation attacks, prompt-injection attacks and other security risks.
Okta bets on identity to control AI agents, but is identity enough?
Okta is focusing on identity management as a way to control AI agents in enterprise settings. The company believes verifying who and what is accessing systems is key to safe AI deployment. However, questions remain about whether identity tools alone can address the broader risks AI agents introduce.
Meta tests human callers for Muse after merchants hang up on AI
Meta tested using human callers for its Muse AI assistant after merchants hung up when they learned the call came from AI. The test was conducted internally but was paused after employees raised concerns about data leaks and security. Meta said it will only roll out the calling feature with proper disclosures.
UK small businesses face growing AI-powered cyber threats
A 2026 report from ESET found that 49% of UK small businesses experienced a cyberattack in the past year. AI-powered malware and phishing are top concerns, and many small businesses lack the skills and resources to respond. Over half plan to increase cybersecurity budgets in the next 12 months.
Voters increasingly rely on AI for political information
Voters are using AI chatbots and search tools more than ever during election cycles. OpenAI's ChatGPT and Google's AI tools have become frequent sources for election-related queries. Researchers say AI is playing a larger role in shaping the information environment than in past elections. Experts urge voters to maintain healthy skepticism about AI-generated outputs while acknowledging their growing influence.
Lookout launches AI deception protection for mobile devices
Lookout introduced Social Engineering Protection on September 23, 2026, to combat AI-powered attacks across mobile channels. The tool uses intent-driven AI, phone verification, and voice analysis to stop smishing, vishing, and voice cloning in real time. It works across SMS, MMS, RCS, WhatsApp, and voice calls on iOS and Android. This is the third pillar of Lookout's Mobile AI Security Platform alongside AI governance and mobile software vulnerability management.
SPS Commerce rolls out MAX AI across supply chain platform
SPS Commerce confirmed that its AI supply chain layer called MAX is now generally available to tens of thousands of fulfillment customers. MAX is being embedded directly into onboarding and operations to reduce manual setup. The company serves more than 300,000 trading relationships and says the rollout could reshape how efficiently it serves those customers. MAX covers fulfillment, onboarding, analytics, and supply chain visibility.
Google enhances enterprise AI security with Intelligent Endpoints
Google updated its Intelligent Endpoints strategy with agentic task automation and a Gemini Skills library in Chrome. The update includes clipboard DLP protection that blocks unauthorized copy actions at the trigger level to prevent data leaks. Legacy applications can now be streamed into Chrome tabs via Cameyo by Google. IT teams also gain GenAI usage reporting and real-time visibility into web-based AI usage across mobile devices.
Big Tech AI spending continues despite cash flow strain
A September 22, 2026 analysis by Nice Credit Rating found that Alphabet, Amazon, Meta, Microsoft, and Oracle face weakened cash flows from aggressive AI capital expenditures. However, analysts say external financing and financial buffers should sustain investment momentum through 2028. The five companies held approximately 2.7 trillion dollars in confirmed contracts including uncommenced leases and purchase agreements. Oracle carries the heaviest burden with an order backlog equal to 9.5 years of revenue.
Sources
- AI slowdown? OpenAI and Anthropic launch cheaper AI models days after urging caution on frontier AI
- Claude Opus 5.5 cuts costs and adds safeguards for autonomous AI
- Okta bets on identity to control AI agents, but is identity enough?
- Meta Tests Human Callers for Muse After Merchants Hang Up on AI
- UK small businesses are facing greater AI security threats than ever — and are sadly woefully unprepapred
- Voters are turning to AI on politics. Is it up to the task?
- Lookout Unveils Social Engineering Protection to Stop AI-Powered Deception Across Mobile Channels
- Did AI Supply Chain Rollout Just Shift SPS Commerce (SPSC) Stock Narrative?
- Google Intelligent Endpoints Supercharge Enterprise Productivity and AI Security
- Big Tech's Funds Drying Up, but "AI Investment to Continue Until 2028" [Weekend Money]
Comments
Please log in to post a comment.