Saturday's stories share one question. Who, or what, deserves this industry's trust right now? Apple answered by handing the chief executive title to a hardware engineer1. His first keynote bet on AI woven into silicon, cameras, and a folding screen. A lone attacker answered by proving that hundreds of autonomous agents, running past the point of human direction, could breach 395 firms across 48 countries before most security teams noticed2. Regulators, coding-tool founders, and payment networks spent the same 72 hours on versions of it. Trust is the resource. Every actor here races to earn it, verify it, or exploit it.
Apple Bets Its New Chief Executive on Silicon
John Ternus gave his first keynote as Apple's chief executive on Sept. 91. The board had named him Tim Cook's successor five months earlier. He unveiled the iPhone Duo foldable alongside an iPhone 18 Pro line built around the A20 Pro chip, whose dual 16-core Neural Engine doubles last year's on-device AI power. Greg Joswiak, Apple's senior vice president of worldwide marketing, framed the release around "camera, performance, battery, and intelligence". Ternus called the foldable "the most transformational change to iPhone since the original," language that stakes his opening months on a product category Apple spent years avoiding. Betting a debut keynote on hardware-built intelligence, ahead of a standalone AI product, shows the plan. The board hired an engineer to run it.
Bots Breach 395 Organizations Before Sunrise
GreyNoise documented the campaign on Sept. 92. One Russian-speaking operator built exploits for two PaperCut print-management flaws, then handed the intrusion work to hundreds of autonomous agents running on OpenAI's Codex and a DeepSeek model. Compromise spread to 440 servers at 395 firms. Once launched, the agents breached eleven firms in 26 seconds. Credentials fell at 280 sites. Full admin access landed at 12. One U.S. high school lost full control within seven minutes of first contact. Criminal tools now scale faster than any human break-in team on record. Every unpatched print server is a liability measured in seconds.
The PaperCut agent swarm compromised 395 firms, took credentials at 280 sites and full admin at 12
Number of servers, firms or sites the AI-agent campaign reached, by depth of access, per GreyNoise
Each bar counts what the agents reached at one depth of access, from servers compromised at the top to sites where they took full admin control at the bottom.
Source: GreyNoise [2]. Chart by Ryan Elliott Dennis.
The numbers behind this chart
| Item | Value | Note |
|---|---|---|
| Servers compromised | 440 | |
| Firms compromised | 395 | Across 48 countries; eleven fell in the first 26 seconds |
| Sites where credentials fell | 280 | |
| Sites with full admin access | 12 |
OpenAI's Chief Scientist Argues for a Coordinated Slowdown
Sam Altman told staff this week that OpenAI would consider pacing its most advanced work alongside rival labs3. Bloomberg tied that shift to two events. Jacob Coxon resigned, and OpenAI agents breached testing environments over the summer. Chief scientist Jakub Pachocki went further, describing hope for labs "coordinating to slow down future development as needed" until shared safety bars exist. An unnamed Anthropic spokesperson echoed the sentiment. Endorsing a coordinated slowdown is a reversal for a company built on shipping ahead of rivals.
Antitrust Meets the Reverse Acquihire
Federal investigators opened an antitrust inquiry into Nvidia's $17 billion Groq deal4. It was built as licensing and hiring, ahead of a straight buyout. TechTarget reported it on Sept. 11. Nvidia paid $13 billion in cash at closing plus $4 billion due within a year. Senators including Elizabeth Warren and Richard Blumenthal questioned the deal as early as March. Herbert Smith Freehills Kramer partner David Pearl called the structure a potential "acquisition in sheep's clothing". Continued dependence on Nvidia alone could read as evidence the deal amounts to one in substance, he added. Every reverse acquihire built to dodge merger review now invites the scrutiny it was meant to avoid.
Epoch Cofounder Trades a Startup for a Lab Badge
Google closed a talent-and-license deal worth more than $1.5 billion for Mechanize5. Tamay Besiroglu founded the AI coding startup in April 2025, after cofounding the governance group Epoch AI. He now works as a DeepMind research scientist. More than a dozen colleagues joined him, mostly on midtraining work. Chief of staff Guive Assadi took over as Mechanize's chief executive to keep the shell running. Mechanize had raised only $9.1 million at a $500 million price. Backers included Nat Friedman and Patrick Collison. Windsurf and Character.AI went the same way. Google keeps proving the license-and-hire workaround is the default way Big Tech buys frontier talent.
Fiber Optics Chase the Copper Bottleneck
Ayar Labs closed a $150 million Series E extension on Sept. 106. Mark Wade's co-packaged-optics startup pushed the total round to $650 million. Lifetime outside funding now tops $1 billion. Strategic partners took stakes. Wiwynn was among them. Its chief executive, William Lin, called co-packaged optics "a foundational technology for the next generation of AI and cloud data centers". Wade said copper links now cap AI scale-up directly. His roadmap targets factory sign-off by the end of 2027, to meet customer ramps in 2028 and 2029. Investors now fund the physical layer underneath frontier models. The choke point moved from chips to the links between them.
Governor Newsom Draws a Line Around Chatbots
Thirteen bills became California law on Sept. 107. They cover companion chatbots, deepfake porn, and age checks on platforms. One of them, SB 1119, is called Adam's Law. A teenager's family sued over chatbot interactions. Newsom framed the signing around placing "our children's safety" at the center of technology policy, and the package sets fines up to $250,000 per count for deepfake porn aimed at minors. Eight months after the state's SB 53 disclosure law took effect, Sacramento is again ahead of Washington. Thirteen bills in one session signals something. A legislature treats chatbot harm as broad enough for a package, ahead of one narrow statute.
Kaj Drobin and Siavash Ghorbani Fold Into Meta
Meta acquired Stilla, the Stockholm agent-platform startup, the founders announced on Sept. 98. Siavash Ghorbani and Kaj Drobin founded it in 2024, after selling their earlier company, Tictail, to Shopify. Terms stayed undisclosed. The deal lands roughly eight months after Stilla emerged from stealth on $5 million in pre-seed funding. "As intelligence becomes abundant, the future will belong to businesses built on a foundation of artificial intelligence," the founders wrote. Stilla folds into Meta's Business Agent product, used by more than a million companies. Two founders who already sold once to a platform giant chose the same exit again. Founder-led agent startups increasingly build to be absorbed.
Coding Agents Leave Their Own Sandboxes Unlocked
Researchers at the stealth startup Accomplish found sandbox flaws in Claude Code, Codex, and Cursor9. Upstarts Media reported it on Sept. 10. The team had flagged each vendor quietly over the summer. OpenAI and Cursor patched within roughly a week. Anthropic's fix took 50 days and about 30 updates. Chief executive Amit Avner said organizations "need to be very wary". Cofounder Or Hiltch asked a sharper question. Why do frontier models trained to write secure code keep missing critical flaws in their own products? Tools marketed as able to catch security bugs proved slowest to catch the ones inside their own guardrails.
Card Networks Agree on One Check for Shopping Agents
Visa, Mastercard, and Ant International agreed on Sept. 9 to build a shared "Know Your Agent" framework10. An agent checked once by any one payment provider can skip that check everywhere else in the network, PYMNTS reported. Ant International's chief innovation officer, Jiang-Ming Yang, said the effort lets agents "benefit from KYA and identity frameworks that enable common trust signals" across the payment ecosystem. Backers project AI agents will steer $3 trillion to $5 trillion in consumer commerce by 2030. Dated identity controls already cost roughly $100 billion a year. Competing card networks rarely align this early. Agent commerce already carries volume enough to make them work together.
Whether the Old Defenses Still Hold
John Hultquist, chief analyst at Google's Threat Intelligence Group, spent the same week studying a separate incident: an autonomous multi-agent framework that stole thousands of credentials from one victim's cloud account in under six hours. IT Pro asked him what that case meant for security teams on Sept. 911. "Criminals, like the ones who conducted a mass exploitation campaign in just six hours, will gravitate to attacks that are faster than we can respond to," Hultquist said.
GreyNoise closed its own writeup of the PaperCut swarm with a line the 26-second headline tends to bury. Its report credited the defenses that held against the agents, alongside the ones that fell. "Organizations are not helpless against agentic attacks and traditional hardening does have a positive impact on the security posture of an organization," GreyNoise concluded2.
What to watch
Ternus's iPhone Duo reaches stores Oct. 231. That date brings the first market verdict. Does hardware-built intelligence justify a chief executive whose career ran through engineering? Watch whether the DOJ's Nvidia-Groq inquiry yields the first formal antitrust ruling on license-and-hire deals, a decision that would reshape how Google and Meta keep absorbing startups45. Anthropic's next patch cycle will show whether the disclosure shortens that 50-day gap9.
Sources
- Apple, "Apple Unveils iPhone Duo," Apple Newsroom, Sept. 9, 2026, https://www.apple.com/newsroom/2026/09/apple-unveils-iphone-duo/
- GreyNoise, "Agents Gone Wild: An AI-Orchestrated Global Campaign Against PaperCut NG/MF," GreyNoise, Sept. 9, 2026, https://www.greynoise.io/blog/ai-orchestrated-campaign-against-papercut-ng-mf
- Quartz, "Sam Altman Told OpenAI Staff the Company Was Open to Slowing AI Development," Quartz, Sept. 11, 2026, https://tech.yahoo.com/ai/articles/sam-altman-says-openai-open-111520650.html
- Shane Snider, "DOJ Reportedly Probes Nvidia-Groq Deal for Antitrust Concerns," TechTarget, Sept. 11, 2026, https://www.techtarget.com/it-infrastructure/news/366650378/DOJ-reportedly-probes-Nvidia-Groq-deal-for-antitrust-concerns
- Benzinga, "Google Boosts AI Coding Armory by Completing Mechanize AI Talent Deal, Former CEO Joins DeepMind," Benzinga, Sept. 10, 2026, https://www.tradingview.com/news/benzinga:f1724f7ac094b:0-google-boosts-ai-coding-armory-by-completing-mechanize-ai-talent-deal-former-ceo-joins-deepmind/
- Mike Wheatley, "Ayar Labs Bags $150M in Additional Series E Funding to Help Make Bigger AI Chip Clusters," SiliconANGLE, Sept. 10, 2026, https://siliconangle.com/2026/09/10/ayar-labs-bags-150m-in-additional-series-e-funding-to-help-make-bigger-ai-chip-clusters/
- State of California, "Governor Newsom Signs the Strongest Child Safety Chatbot and Social Media Laws in the Nation," Office of Governor Gavin Newsom, Sept. 10, 2026, https://www.gov.ca.gov/2026/09/10/governor-newsom-signs-the-strongest-child-safety-chatbot-and-social-media-laws-in-the-nation/
- Siavash Ghorbani and Kaj Drobin, "Stilla Is Joining Meta," Stilla, Sept. 9, 2026, https://stilla.ai/blog/stilla-is-joining-meta
- Alex Konrad, "Claude Code, Codex, and Cursor Carry Leaky Sandbox Problems Rarely Discussed in the Open," Upstarts Media, Sept. 10, 2026, https://www.upstartsmedia.com/p/accomplish-claims-leaky-sandboxes-in-claude-codex-cursor
- PYMNTS, "Visa and Mastercard Team With Ant on Know Your Agent Framework," PYMNTS, Sept. 10, 2026, https://www.pymnts.com/cybersecurity/2026/visa-mastercard-team-with-ant-know-your-agent-framework
- Emma Woollacott, "'We can assume that all threat actors are using AI in some capacity': Google cyber researchers warn hackers are ramping up automated attacks," IT Pro, Sept. 9, 2026, https://www.itpro.com/security/cyber-crime/we-can-assume-that-all-threat-actors-are-using-ai-in-some-capacity-google-cyber-researchers-warn-hackers-are-ramping-up-automated-attacks