AI Decodes a 217-Year-Old Napoleonic Military Letter in 6 Hours
Carter Church, an engineer at the cybersecurity company SentinelOne, used AI to decipher a Napoleonic-era military message that had gone unsolved for more than two centuries.
- The encrypted document was sent to French General Auguste de Marmont on Napoleon’s orders as Europe moved toward another major conflict, though it was not written directly by Napoleon.
- OpenAI’s GPT-6 Astra reportedly handled the process end to end, first transcribing the image and identifying repeated symbols before analyzing the document’s homophonic substitution cipher.
- The model also wrote code and cross-referenced historical records, completing the decipherment in roughly six hours.
- The letter revealed details about the positions of the Bavarian, Polish and French forces, as well as references to the movements of the Russian and Austrian troops.
- Using the decoded troop numbers, officers’ titles and unit positions, Church re-dated the document to late March 1809, about two weeks before Austrian forces launched its offensive against France.
Several researchers had previously attempted to reconstruct the letter, but no complete decipherment had been published before this effort. Church says Satoshi Tomokiyo, the Tokyo-based researcher who maintains Cryptiana, reviewed the solution and marked the entry as “solved.”
Trump Announces Formation of AI “Super Intelligence Force”
President Trump announced Sunday the formation of a “Super Intelligence Force” that will coordinate federal efforts around AI, with the goal of ensuring the U.S. remains the global leader in “super intelligence.”
- Jay Clayton will reportedly chair the task force, with FTC Chair Andrew Ferguson, Undersecretary of War for Research and Engineering Emil Michael, and OPM Director Scott Kupor serving as vice chairs.
- The group will have 120 days to produce a report examining the opportunities and risks created by AI and developing plans for responding to potential threats.
- The task force is expected to “develop plans for responding to SI-enabled threats to our society, while preventing overregulation and regulatory capture that would stifle innovation and competition.”
- Clayton has argued that the U.S. should not pull back from AI development as China advances, saying that failing to remain ahead could itself be one of the biggest risks.
Trump first floated the idea in September, saying he would form an AI task force in response to the growing debate over AI safety. Sunday’s announcement delivers on that promise, with the new name also being part of the administration’s recent push to rebrand AI as “super intelligence.”
You can now virtually try on clothes with ChatGPT
ChatGPT can help you find the perfect outfit for a meeting or event, and now it goes a step further by showing you how those clothes might look on you.
Step 1: Open ChatGPT and search for the clothes and accessories you are looking for.
Step 2: Hover over one of the results and click the Try on button that appears.
Step 3: Upload a selfie or a full-body photo to see how the clothing or accessory might look on you.
Step 4: ChatGPT saves your reference photo for future try-ons. You can add, replace, or delete it at any time by going to Settings > Personalization > Reference photos.
Step 5: Hover over an item and click the bookmark icon to add it to Favorites. Saved products are stored in the ChatGPT Library, so you can come back to them whenever you like.
Notes
- These features are available in ChatGPT on both mobile and web.
- It is powered by the newly launched ChatGPT Images 2.5 model.
- You are not limited to the items ChatGPT finds for you. You can also upload a picture of any clothing or accessory and simply ask ChatGPT to show how it could look on you.
Florida woman arrested following conversation with Claude
A Florida woman has been arrested after allegedly making violent threats against the Lee County Sheriff’s Office in a conversation with Claude. The chatbot’s safety systems flagged the messages, which were escalated to a human review team that then reported them to law enforcement.
OpenAI expands visual ads in ChatGPT
The company is introducing a new visual ad format in ChatGPT that will initially appear during image generation, showing sponsored products and services separately from the images users create. The ads will be tested with a group of U.S. advertisers later this month.
Musk says SpaceXAI will be renamed SpaceXSI
Elon Musk confirmed the change on X but gave no timeline. It would be the company’s second rename this year, after SpaceX took over xAI in February and the AI business became SpaceXAI in July.
Meta’s AI is profiling your friends
Meta’s new Muse AI assistant can book tickets, sell items, and handle other tasks, but it also builds detailed profiles of people in users’ personal networks, raising fresh privacy concerns.
Muse encourages users to share sensitive information including emails, banking details, and passport data, while critics warn that combining those records with relationship profiles could expand surveillance and misuse risks.
Meta already patched a serious vulnerability discovered during Muse’s early rollout, making the trade-off hard to ignore: smarter assistants become more useful as they learn increasingly personal details.
OpenAI safety lead quits over “broken” culture
David Robinson quit OpenAI after 3.5 years and announced his exit in The Atlantic, calling the company’s culture “broken.” By his account, he oversaw safety reports for 12 frontier launches and drafted its Preparedness Framework.
Robinson argues AI labs should run like nuclear plants and airports, with layers of redundancy against human error. Instead, he says his team sprinted so hard they rarely had time to consider big changes.
Robinson’s exit follows OpenAI cutting three safety staffers it says mishandled sensitive information. Back in 2024, departing safety co-lead Jan Leike warned safety had taken a backseat to shiny products. Same alarm, new messenger.
Anthropic seeks religious guidance for Claude
Anthropic co-founder Chris Olah spent the past year consulting religious scholars from multiple faiths, asking them to examine Claude’s possible consciousness and help shape its morals, according to The New York Times.
Scholars joined NDA-bound seminars around Anthropic’s 84-page “Soul Doc,” while Olah also discussed AI consciousness with Pope Leo XIV, reportedly objecting when a papal letter dismissed the possibility.
The split is getting unusually philosophical: Pope Leo says algorithms lack humanity’s “spark,” while Sam Altman warns against giving AI religious force and Anthropic keeps exploring whether Claude might deserve moral consideration.
Trump launches Super Intelligence Force
President Trump announced a new Super Intelligence Force on Truth Social to coordinate federal AI efforts. The Wall Street Journal reports that Director of National Intelligence Jay Clayton will chair it.
The name follows Trump’s executive order directing federal agencies to swap “AI” for “Super Intelligence” in official documents. Trump also promised an AI Force and czar, and later praised Clayton for the role.
Its charter reportedly pairs SI threat planning with preventing overregulation, and Clayton told the WSJ one of the biggest risks is “not being first.” Safety made the agenda, but so did speed.
Sam Altman says society should accept some AI risks
In an interview with Politico’s new Decoded newsletter, Sam Altman said people should tolerate “some bad things” from AI, like hacks and scams, because, he argued, the benefits outweigh those harms.
Altman draws the line at extreme risks like losing control of AI. He also rejected the idea of one lab holding the technology and sharing its gains, a view Anthropic says covers only frontier models.
Three weeks ago, Altman backed Anthropic CEO Dario Amodei’s call to pace the frontier and pledged outside evaluators employee-like access. Agreeing on oversight, it turns out, isn’t the same as agreeing on rules.