The first Claude chat came on August 11 in San Antonio. Nathaniel Michael Carrasco, 22, told Anthropic AI he wanted to get a gun and shoot students at Serna Elementary School, according to an arrest affidavit. The FBI’s National Threat Operations Section notified local police on August 28. Officers linked Carrasco to the account through emergency disclosure requests to T-Mobile, Charter Communications, and Google that matched his phone number, IP traffic, and address to the account.

Claude chats lead police to offenders

Carrasco was arrested on August 30, the school’s principal said, on a third-degree felony charge of terroristic threat causing public fear of serious bodily injury. The school only discovered the specific threat against its campus through news coverage, she said. Three days after the San Antonio chat, on August 14, a man allegedly told Claude he had bought an AR-15 and had Amodei “in his sights,” according to a San Francisco police report. The following Tuesday, Anthropic told the SFPD.

Officers established that the man was not at the company’s office at 500 Howard Street. An Anthropic spokesperson said the company banned the account and referred the case to law enforcement, saying it was the safeguards process working as intended. As of early September, he had not been arrested or charged. Anthropic’s privacy policy, which went into effect September 10, says it can share data with law enforcement if it has a good-faith belief it is reasonably necessary to prevent serious harm.

Its consumer terms reserve the right to report a user’s inputs, outputs, or actions to law enforcement “in its sole discretion.” The arrest report said Carli Michelle Heller, of Bonita Springs, Florida, wrote on September 26 that she was going to “shoot up” the Lee County Sheriff’s Office. Investigators say she wrote the next day that she obtained a new gun. The report says the statements were flagged by Anthropic’s monitoring. A human review team referred them to law enforcement owing to their severity.

Deputies detained Heller at her home without incident. She faces a charge of making a written threat of violence and has a court date in November. Sheriff Carmine Marceno said she later described using AI like a diary. “Artificial intelligence is a powerful tool, and like any technology, it can be misused,” Marceno said in a statement. In April, Cryptopolitan reported that Anthropic started requiring a government ID and a live selfie for some features. In August, Cryptopolitan reported that the company set Claude’s memory to skip health, politics, and gender identity unless users opt in.

The post Claude chats intercepted by police in seven weeks first appeared on Coinfea.

A September report by Anthropic named seven China-based firms that were reportedly tapping Claude through mass unauthorized access. Besides DeepSeek and Moonshot, that list included Alibaba, Zhipu (also known as Z.ai), SenseTime, MiniMax, and Xiaomi. After the report landed, the Cyberspace Administration of China called in representatives from all seven companies for several rounds of questioning, and officials are now trying to decide if the alleged conduct broke China’s rules on moving data across its borders.

Anthropic fingers DeepSeek for tapping Claude for answers

Anthropic’s own September 2026 threat intelligence report covers activity that the company said it detected and shut down between December 2025 and August 2026. Anthropic accused DeepSeek of relaying selected user requests to Claude while leading those users to believe they were dealing with DeepSeek’s models. Some of the forwarded material was sensitive, including information tied to a Chinese technology firm’s AI program, data linked to a Russian government agency, and records from a Chinese municipal police system.

Anthropic said DeepSeek extracted Claude’s hidden reasoning steps, known as its “thinking signature,” and then used that stolen knowledge to help train its own systems. Notably, during the July period Anthropic examined, DeepSeek rerouted more than 12.1 million of its users’ queries to Claude in just 14 days. Anthropic also revealed that Moonshot AI silently forwarded almost 300,000 customer requests to Claude across a 10-day stretch. The requests were sent through 5,380 fraudulent accounts that mostly appeared to sit in Singapore and Japan.

Anthropic added that it could not confirm whether Moonshot told customers that their requests were going to be sent to a third party. In an earlier account of the report, Cryptopolitan noted one Moonshot-linked user whom Anthropic judged likely affiliated with the Chinese military, using the service to review surveillance footage. Anthropic has previously accused several AI companies of “distillation,” a routine technique in which a smaller “student” model learns from a larger “teacher” model’s outputs.

Anthropic says it has identified and disrupted such efforts at seven Chinese labs since February 2026. Alibaba was the heaviest user of the seven, with more than 151 million exchanges between May and July 2026, a figure Cryptopolitan reported earlier this month. Beijing has pushed back on the distillation narrative, with its officials framing Washington’s complaints as an effort to hold back China’s AI sector. Meanwhile, U.S. and Chinese officials recently agreed to keep talking on AI safety and emergency communication, with another meeting expected in Shenzhen within two months.

The post DeepSeek and Moonshot under investigation over data routed to Claude first appeared on Coinfea.

In the second quarter of its 2026 fiscal year, AMD reported $11.5 billion in revenue, up 50% from a year earlier. Data center products accounted for $6.7 billion of that total revenue. Earnings grew even faster, with AMD reporting earnings per share of $1.66, an 82% increase from the same period a year earlier. The company has also indicated that it plans to raise prices by 10%, which could give it another potential boost to future revenue.

AMD joins other firms as investors continue to bet on AI-related demand

The results came in as demand for AI chips continued to dominate the semiconductor market. AMD is simply the latest chipmaker to cross the $1 trillion mark as investors continue betting on AI-related demand. A newly announced partnership with Anthropic has also put AMD further in the spotlight. The chipmaker’s valuation is increasingly tied to AI workloads, so landing a deal with the Claude maker gives investors another signal of potential demand.

AMD is having a very great year, with its stock up more than 180% in 2026. The chipmaker remains far behind AI leader Nvidia, though, a company with a market value of about $5.4 trillion. The company’s shares traded around $610 after gaining 8.99% over the previous week, reaching $610.12 and an intraday high of $613.92. Weekly volatility for the stock was at 11.73%, and the relative strength index hit 65.51, bringing it close to overbought territory.

The stock is expected to consolidate over the next five sessions, with the forecast putting it between $562.41 and $636.4, and the $559.81 price point seen as a key support level. Viktoras Karapetjanc, an expert at Traders Union, expects the rally to continue, saying AMD could maintain its bullish momentum as long as earnings growth and demand for AI chips remain strong. AMD CEO Lisa Su said on the company’s earnings call last month that data center sales were expected to double in 2027.

The post AMD rockets to $1 trillion market cap on AI chip demand first appeared on Coinfea.

The tool they’re using is Claude, which can read, reason, and follow detailed rules like a real employee. Marco Argenti, Goldman’s tech chief, said they’re still testing the agents but plan to roll them out soon. He said these bots will speed up tasks that normally take forever, like reconciling transactions or going through compliance paperwork. “Think of it as a digital co-worker for many of the professions within the firm that are scaled, are complex, and very process-intensive,” Marco said. That’s not theory. It’s already being tested inside the bank.

Goldman set to expand Claude’s role after early testing

Goldman actually started with a coding bot called Devin. It worked well enough for engineers, but Marco said they quickly noticed Claude was better at more than just code. The team tested Claude on accounting and compliance jobs and was caught off guard when it actually handled them. “Claude is really good at coding,” Marco said. “Is that because coding is kind of special, or is it about the model’s ability to reason through complex problems, step by step, applying logic?”

They tried it on massive documents, hard rules, and messy spreadsheets. Claude was able to understand the rules, apply judgment, and finish the job. Marco said the team was surprised by how strong it was in areas like compliance, not just tech. That’s when they decided to use Anthropic’s model on trade reconciliation and client vetting, too.

David Solomon, the CEO, is already turning the whole bank toward AI. He announced last year that Goldman is starting a long-term plan to bring in generative AI across the company. This isn’t about experimenting. It’s about cutting down on new hires and doing more with fewer people. That includes using AI to run operations without needing armies of junior staff. The plan is simple: do more, hire less.

Marco said they aren’t firing anyone yet. He called it “premature” to assume jobs will disappear. But he did say that third-party contractors might get cut out as the AI gets stronger. That means companies that help Goldman with compliance or accounting might not be needed anymore. And Claude isn’t done. Marco said the next use cases might include internal surveillance or making pitchbooks for deals.

The post Goldman Sachs partners with Claude to develop AI agents for banking tasks first appeared on Coinfea.

“Claude Code has become the go-to choice for coders everywhere, and so it was no surprise to learn OpenAI’s technical staff were also using our coding tools ahead of the launch of GPT-5. Unfortunately, this is a direct violation of our terms of service,” Christopher Nulty, an Anthropic spokesperson, said in a statement.

Under its commercial terms, Anthropic specifically states that customers may not develop another product or service, create their AI models using Anthropic’s platform, or attempt to dissect or copy its technology. This adjustment to Claude’s availability coincides with OpenAI’s internal preparations for GPT-5, the forthcoming model touted for enhanced coding capabilities.

Anthropic announces weekly usage cap on Claude codes

According to insiders, OpenAI integrated Claude into its proprietary toolchain via custom APIs rather than the usual chat portal. This arrangement enabled OpenAI to assess Claude’s performance on programming and writing tasks, and to evaluate its responses to safety-related queries regarding areas such as CSAM, self-harm, and defamatory content. The findings were then used to measure Claude against the in-house models of OpenAI and to tune OpenAI’s system accordingly.

Responding to the development, Hannah Wong, chief communications officer at OpenAI, said, “It’s industry standard to evaluate other AI systems to benchmark progress and improve safety. While we respect Anthropic’s decision to cut off our API access, it’s disappointing considering our API remains available to them.”

Nulty also added that granting API access for purposes like benchmarking and safety reviews is still important for the industry. The company is yet to clarify how the current restrictions on Claude might affect its ongoing evaluation. API restrictions have long been wielded as leverage in the tech sector. For instance, Facebook previously cut off Vine (owned by Twitter), and more recently, Salesforce imposed constraints on rival Slack integrations.

Anthropic recently revealed plans to introduce weekly caps on Claude Code to reduce continuous background consumption by users. The company said in an email and on X that the new limits will start on August 28. These restrictions will affect users on the $20 Pro tier as well as those on the $100 and $200 Max tiers. Existing five-hourly usage quotas will remain unchanged.

In addition, the update is expected to include two separate weekly quotas: a general usage cap and a model-specific allowance for Claude Opus 4. Max-tier subscribers will have the option to purchase additional capacity at Anthropic’s standard API pricing once they exceed the weekly thresholds. This adjustment builds on a subdued deployment of usage ceilings for Claude Code implemented earlier this summer.

The post Anthropic cuts OpenAI’s access to its Claude API first appeared on Coinfea.