Siri Reboots, Apple Sues OpenAI, Russia Owns Your Router

Apple Ships A Siri That Works, Then Sues Its Rival

The iOS 27 public beta puts a rebuilt Siri at the center of the iPhone, and by the reckoning of the first testers out of the gate, it is finally the assistant Apple promised at WWDC roughly three product cycles ago. Wired frames it as a genuine overhaul. The Verge's tester says a week with the new Siri has already rearranged how they touch their phone, which is either a triumph of design or a confession about how bad the old one was. Probably both.

On the same news cycle, Apple filed suit against OpenAI, accusing Sam Altman's company of obtaining confidential hardware samples through the recruitment of Apple engineers. The complaint reads as a warning shot aimed at the Jony Ive collaboration, the mysterious OpenAI device that has been leaking vibes for a year. Apple's argument is that OpenAI did not just hire people, it hired them to smuggle prototypes out the door. If any of that survives discovery, the AI hardware race gets its first real courtroom season.

The timing is not subtle. Apple wants the story to be that Siri is good now and that OpenAI is a thief. Whether both are true matters less than whether both stick.

Your Router Is A Russian Asset

CISA issued a warning that Russian state-backed operators are actively compromising residential and small-office routers in the United States. The pattern is familiar and grim. Consumer routers run outdated firmware, ship with default credentials, and sit unpatched for years because nobody thinks of them as computers. Attackers use them as staging infrastructure, obfuscation layers, and occasionally as footholds into the networks behind them.

The practical read is that the perimeter of a modern household is a plastic box from 2019 that the ISP mailed you. Treat it accordingly. Firmware updates, disabled remote administration, and a healthy suspicion that the light blinking in the corner of your living room is doing work for someone in Moscow.

On the defensive side, Ars Technica documents a novel tactic where security teams are firing prompt injection at hostile AI agents to stop them mid-action. Context bombing, as the practitioners call it, floods an agent with instructions designed to trip its safety heuristics or send it into a refusal loop. It is the security equivalent of throwing sand in the gears, and it works because the same fragility that makes agents dangerous to deploy also makes them easy to derail. Offense and defense are now using the identical exploit class. Whichever side has better prompts wins.

Research, Retaliation, And A Baseball Game

MIT Tech Review takes a careful look at Anthropic's latest interpretability research and lands on the honest verdict, which is that the results are genuinely interesting and do not tell us whether frontier models are safe. This is the shape of the field now. Real progress on understanding what these systems are doing internally, with the humility to admit that understanding is not control.

The New York Times covers a new documentary about the eBay cyberstalking case, in which employees of a Fortune 500 company mounted a harassment campaign against two journalists who had written critically about them. Live cockroaches, a fetal pig, surveillance. The film is a reminder that the worst online abuse does not require AI, just a corporate expense account and a security team that has lost the plot. It reads as prologue to the automated harassment era we are now entering.

And because the universe demands balance, a Brooklyn teacher named Lindsay Barnett successfully brought Backyard Baseball back to market. Pablo Sanchez lives. It is the only story in the slot with no villain, and we are noting it for the record.

The Read

The pattern this cycle is convergence. Apple is fighting OpenAI in court while shipping the assistant it needs to compete with them. State hackers and AI defenders are exploiting the same prompt-level fragilities from opposite sides. The interesting research keeps landing without resolving anything. Nothing here is catastrophic on its own. All of it is the texture of a landscape hardening into something adversarial.