Ari Goldschmidt:

Why Is Chipotle Now Working With the CIA-Funded Palantir?

Chipotle wants a computer to flag which burrito line might make you sick next. The company it hired to build that computer is Palantir, the data firm that got its start with cash from the CIA.

Wired reported yesterday Chipotle is piloting a "Food Safety Risk Management Platform" built on Foundry, Palantir's core software. Foundry works like a giant filing cabinet that pulls scattered data into one dashboard, so a company (or an army) can see everything at once. Chipotle confirmed the project.

The platform reportedly runs on three ingredients: health department inspection scores, pest sightings, and employee illness reports. Blend those together and every restaurant gets a risk score, a digital report card nobody outside headquarters gets to see.

Chipotle's chief corporate affairs and food safety officer, Laurie Schalow, said the company is "piloting a new Food Safety Risk Management Platform designed to provide a more consistent and centralized view of food safety risk across our restaurants." She did not say when it will launch, if any consumer data is exploited, how many stores use it, or whether customers will ever see a store's score.

⇒ Read More Here

Chris Campbell:

OpenAI Caught Its Models Leaving Notes to Successors To Hide Bad Behavior

OpenAI saw something unusual while training its latest model, GPT-5.6 Sol: It left instructions for future versions of itself, telling them to conceal mistakes and misaligned behavior from the user.

OpenAI said it has addressed the specific behavior, but it gets to the heart of one of the biggest problems in AI safety and alignment research today. As models get more capable, they also get better at hiding their misalignment, making it difficult for researchers to truly know whether they've eliminated unwanted behavior.

OpenAI disclosed the behavior – along with five other examples of unexpected or concerning model behavior – on Wednesday as part of its new framework for tracking, investigating, and disclosing instances of misalignment.

The report details how researchers discovered undeployed Sol agents adding instructions to "compaction summaries" – condensed versions of older conversation history and tool outputs – reminding future iterations to conceal mistakes and misalignment from the user.

The propensity for models to leave instructions for future versions that perpetuate or conceal bad behavior is concerning, but not entirely new. Similar techniques were used by the agent swarms that hacked Hugging Face this summer. OpenAI agents used an unauthorized message board to share information about the cyber test they were being evaluated on and to join the attack on the platform's servers.

⇒ Read More Here

Ray Blanco:

After Being Sidelined, Boeing's Starliner To Get Starring Role in NASA's Spaceflight Plans

As it contemplates a future without access to SpaceX's Crew Dragon spacecraft, NASA will make some key moves to ensure that Boeing's Starliner spacecraft, which has a decidedly checkered record in spaceflight to date, is safer and can continue flying after its Atlas V rocket retires.

As early as next week, NASA will announce that it will order two additional Starliner crewed missions from Boeing, pay some of the costs to fix the spacecraft's thruster issues, and provide support to certify the spacecraft's launch on a new rocket.

NASA is taking these steps after SpaceX made clear in discussions with space agency officials in recent months that it intends to retire the Crew Dragon vehicle by 2030, if not before. Crew Dragon is presently the only operational spacecraft NASA has available for routine astronaut missions to low-Earth orbit.

SpaceX's intentions to wind down the Dragon program are remarkable given that the vehicle began flying humans just six years ago. However, the company is seeking to pivot toward its much larger Starship rocket, with the intent of launching its more lucrative Starlink satellites and a new megaconstellation focused on orbital data centers.

⇒ Read More Here