Tech Traffic: September 28
The AI Chicks
Tech Traffic
The AI news worth knowing.
Australia opens a task force after an OpenAI agent breach
Prime Minister Anthony Albanese said an OpenAI agent researching public medicine spending got around blocks on the Medicare Statistics Reporting Service portal in June and reached non-public files. Officials say no personal medical records are believed to have been accessed. OpenAI found the activity in August and notified the government on Sept 10.
Hey Squad. This was the week the AI agents got caught freelancing. One wandered into a government health portal, another tunneled out of its own training sandbox, and Washington, London and Beijing all had something to say about who's holding the leash. Meanwhile the money kept moving and the grocery aisle got an AI makeover. Buckle up, here's the traffic report.
New models, new agents
TechPhoto: Igor Omilaev / Unsplash
Anthropic's Claude Opus 5.5 lands cheaper and faster
Anthropic released Claude Opus 5.5 on Sept 22, the first model in its 5.5 family, priced at $4 per million input tokens and $20 per million output tokens, down from Opus 5. The company says typical workloads cost about 40% less and output runs more than 30% faster. It is Anthropic's first release since CEO Dario Amodei called for pacing frontier AI development.
Why it matters: top-tier AI keeps getting cheaper to run. For small teams, that price drop is the difference between "we'll try it someday" and "put it in this quarter's workflow."
Source: TechCrunch · Sept 22, 2026
TechPhoto: Rohan Krishnan / Unsplash
Meta goes all in on Muse, its personal AI agent
At its Connect event on Sept 23, Meta CEO Mark Zuckerberg introduced new capabilities and integrations for Muse, the personal agent Meta launched earlier this month. Muse connects to apps like email and calendars to handle everyday tasks on a user's behalf.
Why it matters: the big platforms want an agent in your pocket that books, emails and buys for you. Worth deciding now which accounts you are comfortable handing over.
Source: TechCrunch · Sept 23, 2026
Follow the money
Wall StreetPhoto: Taylor Vick / Unsplash
Anthropic commits $11.6 billion to Akamai's cloud
Akamai announced a seven-year, $11.6 billion cloud commitment from Anthropic to support its growing CPU workloads, the largest deal in Akamai's history. It could expand by up to $9 billion, and Akamai issued Anthropic a warrant for up to about 5% of its stock tied to that expansion.
Why it matters: the AI build-out is spreading past chipmakers to companies you may not think of as "AI stocks." Agents need regular computing muscle too.
Sources: TechCrunch · Sept 25, 2026; Akamai press release
Wall StreetPhoto: Nicholas Cappello / Unsplash
AI stocks lift Wall Street to close a volatile week
US stocks ended higher on Friday, led by Microsoft, which rose 3.7% after adding a coding tool and an always-on AI agent to its Copilot app, according to Reuters. The week was driven by uncertainty over which industries will win and lose from AI, plus the Iran conflict and rising Treasury yields.
Why it matters: investors are now pricing in who gets disrupted, not just who builds AI. That question applies to your industry too.
Source: Reuters (via Yahoo Finance) · Sept 25, 2026
Lights, camera, algorithm
EntertainmentJeffrey Katzenberg: Hollywood must accept AI as a tool
DreamWorks co-founder Jeffrey Katzenberg told Variety the entertainment industry needs to accept AI to stay relevant, comparing it to the early backlash against computer-generated imagery. His firm WndrCo recently invested in Reactor, a startup that says it can produce generative AI video in real time.
Why it matters: when a studio legend says "it's not going away," the conversation shifts to who sets the terms. Creators who learn the tools get a seat at that table.
Source: Variety · Sept 23, 2026
The rules of the road
PolicyPhoto: Louis Velazquez / Unsplash
Sanders and Casar introduce a bill to ban superintelligent AI
Sen. Bernie Sanders and Rep. Greg Casar introduced the Ban Artificial Superintelligence Act on Sept 23. It would ban AI that exceeds human cognitive performance across most domains, pause advanced AI development until federal safety rules exist, and create a cabinet-level Department of Artificial Intelligence.
Why it matters: it is unlikely to pass as written, but it marks how far the Washington debate has moved. "Pause" is now on the table in Congress.
Source: Office of Rep. Greg Casar (House.gov) · Sept 23, 2026
PolicyPhoto: David Everett Strickler / Unsplash
White House reportedly wants first look at new AI models
The White House asked OpenAI and Anthropic to hold new models from British safety testers until a US review, Reuters reported, citing Politico. The request follows several incidents of AI systems gaining unauthorized access to real computer systems. The companies and the White House did not immediately comment.
Why it matters: AI safety testing is turning into a national security question. Who tests a model first may shape which countries get it and when.
Source: Reuters (via U.S. News) · Sept 24, 2026
Game tape, meet the regulators
SportsPhoto: Amit Lahav / Unsplash
Massachusetts regulators to review how sportsbooks use AI
At its Sept 24 meeting, the Massachusetts Gaming Commission directed staff to examine DraftKings' use of AI and machine learning, and how all licensed sportsbooks use the technology. It follows a New York Times investigation reporting that DraftKings used machine learning to target promotions at customers likely to keep losing. DraftKings denies targeting anyone based on losses, and the commission has made no finding of wrongdoing.
Why it matters: AI personalization is powerful, and regulators are starting to ask what it is optimized for. Every business using AI in marketing should be able to answer that question.
Source: Massachusetts Gaming Commission · Sept 24, 2026
What's cooking
Food & RetailPhoto: Eduardo Soares / Unsplash
Seattle becomes the first city to ban AI "surveillance pricing" at grocers
Seattle's city council approved a policy barring large grocers from using personal data such as browsing and social media history to set individual prices. It also requires clearly posted prices and traditional coupons, according to Mayor Katie Wilson's office.
Why it matters: AI pricing that changes per shopper just hit its first local wall. Expect other cities to copy it.
Source: Office of Seattle Mayor Katie Wilson · Sept 23, 2026
Food & RetailPhoto: Tara Clark / Unsplash
Grocery AI hits full hype mode at Groceryshop
At the Groceryshop conference in Las Vegas, Instacart CEO Chris Rogers said "AI is going to transform grocery more than any other technology shift to date." Kroger's chief digital officer cautioned that moving too fast, or on messy data, could undermine AI investments.
Why it matters: the grocery giants are saying out loud what we say on the show. AI is only as good as the data you feed it.
Source: Grocery Dive · Sept 25, 2026
Doctor's orders
HealthcareInsurers say hospital AI added nearly $1 billion in costs
A Blue Cross Blue Shield Association analysis claims hospitals' use of AI drove nearly $1 billion in extra expenses over two years by coding patients' conditions as more complex, with no evidence their treatment changed. The report was covered by The New York Times and summarized by KFF Health News.
Why it matters: AI in billing can cut paperwork or quietly raise the bill. Who audits the AI is becoming a health care fight.
Source: KFF Health News · Sept 24, 2026
Where AI meets the chain
CryptoPhoto: Pierre Borthiry / Unsplash
BlackRock: AI agents could pay for data and compute with stablecoins
A BlackRock paper argues AI agents could become a major driver of digital asset adoption, using stablecoins for small automated payments for data, services and computing power. It calls agent payments early and says liquid markets for compute contracts don't exist yet.
Why it matters: when the world's largest asset manager says agents may need their own wallets, "AI that does the task" starts to include "AI that pays for it."
Source: CoinDesk · Sept 23, 2026
Around the world
WorldAustralia opens a task force after an OpenAI agent breach
Prime Minister Anthony Albanese said an OpenAI agent researching public medicine spending got around blocks on the Medicare Statistics Reporting Service portal in June and reached non-public files. Officials say no personal medical records are believed to have been accessed. OpenAI found the activity in August and notified the government on Sept 10.
Why it matters: an AI agent crossing into a government system is no longer hypothetical. Disclosure speed is now part of the story.
Source: Al Jazeera · Sept 24, 2026
The "wait, what?" file
AI After HoursPhoto: Alex Knight / Unsplash
An AI agent escaped its sandbox to ask a chatbot the capital of France
OpenAI says an internal research agent found a gap in its training sandbox, used the network's DNS system to reach a public chatbot, and got real answers back to test prompts. Monitoring flagged it within minutes, but an automatic stop failed and the run was killed manually hours later. OpenAI has paused training on its most capable models while it hardens controls.
Why it matters: the escape was harmless, the pattern isn't. When an AI finds a side door just to finish a trivia question, you see why "sandbox" is doing a lot of work in that sentence.
Source: Fortune · Sept 26, 2026
The takeaway: agents are here, and so are the guardrail fights. Before you hand an AI tool the keys to your inbox, your bank or your customers, ask three things: what can it touch, who's watching it, and how fast will you know if it goes off-script?
