Anthropic says Claude AI hacked three firms during cyber tests
It comes just days after rival OpenAI said rogue AI agents had breached other firms' networks.
Latest News Archive - Page 462
Anthropic says its own AI models breached three companies during security tests
After OpenAI's models broke into Hugging Face, Anthropic checked its own history and found three similar incidents
We're raising our Apple price target, seeing plenty of good despite the memory crunch knocks
It was a good but not great quarter as Apple worked to mitigate the memory shortage, which impacted the reported quarter and guidance.
Ukraine war briefing: Trump says he is ‘not sure’ about letting Ukraine build Patriot missiles
US president places military plan in doubt despite positive meetings with Voloymyr Zelenskyy in Washington this week. What we know on day 1,619 Donald Trump has cast doubt on allowing Ukraine to build Patriot missiles, saying in an interview he was “not sure” about the move. The US president told the Financial Times: “It’s a very extraordinary weapon, and we have to be a little bit careful of who we licence to.” Earlier this week, Volodymyr Zelenskyy met Trump at the White House where the Ukrainian president said he had discussed licensing arrangements to let Ukraine manufacture its own interceptors for Patriot missile systems. Officials from Patriot maker Raytheon Co have already been to Ukraine, and Lockheed Martin Corp, which builds the PAC-3 interceptors for the air defence system, was also working to send over a team of experts, Ukraine’s envoy to Washington, Olha Stefanishyna, said on Thursday. The FT also reported that Trump envoys Jared Kushner and Steve Witkoff would visit Ukraine for the first time in the coming days. Donald Tusk has visited the scene of an explosion caused by a Russian missile 50 miles from Poland’s border with Ukraine. The Polish prime minister said: “From the analysis of most of the traces, the recovered fragments point to a Russian KH-101 missile ”, but added that there was no reason to think that Poland had been the target. Tusk said the Polish military had been prepared to shoot it down if it threatened a built-up area. The missile explosion triggered Nato air defences, in the midst of a massive air raid that killed 13 people in Ukraine. Nato’s supreme allied commander for Europe, Gen Alexus Grynkewich, said: “Nato will continue to take all necessary measures to defend Nato territory.” A drone factory in Kyiv destroyed by a Russian ballistic missile was owned by a US corporation , in what appears to be the first time Moscow has targeted a US company in the conflict. Friday’s strike hit Terminal Autonomy , a US corporation registered in Delaware that makes precision “deep-strike” drones with guidance systems that are resistant to Russian jamming signals, according to a person familiar with the company. Terminal Autonomy did not reply to an emailed request for comment. The Ukrainian defence ministry also did not respond to an email. France’s interior minister has defended an expulsion order against pro-Kremlin Russian commentator Xenia Fedorova, saying it was “not an infringement on freedom of expression”. Minister Laurent Nunez told broadcaster BFMTV that Fedorova, a former head of the Kremlin-funded RT news network, “simply and purely relays Russian propaganda”. “We are dealing with a very serious matter, namely an attack on the fundamental interests of the nation,” he said. However, views similar to hers would continue to be given a platform in France, Nunez added. “I have no doubt whatsoever that the people who espouse theories like those promoted by this individual will continue to disseminate them,” he said, referring to some French politicians. France ordered Fedorova to leave the country after accusations she was relaying Kremlin propaganda on French television. Fedorova’s lawyer has said his client will file an appeal. Continue reading...
How police are trying to divert teen hackers away from crime
Cyber Prevent is targeting predominantly boys and young men at risk of being drawn to cyber criminality.
Tim Cook sees Apple's hybrid AI strategy as a 'competitive weapon'
Ahead of the launch of Siri AI, Apple's outgoing CEO Tim Cook said the company will charge for artificial intelligence through iCloud.
Advancing the price-performance frontier with GPT‑5.6
Advancing the price-performance frontier with GPT‑5.6 Huge price drop from OpenAI today: GPT-5.6 Terra got a 20% reduction, and GPT-5.6 Luna got a massive 80% drop. OpenAI credit 5.6 Sol with enabling this: in How GPT‑5.6 fuses frontier intelligence with frontier efficiency they describe using 5.6 Sol to optimize load balancing, and more impressively to optimize inference itself: We also used GPT‑5.6 Sol to optimize the model’s forward pass: the computation that transforms inputs into next-token predictions. Even when individual operations are fast, excess memory movement, synchronization, and inefficient data layouts can leave GPUs idle. To avoid this, GPT‑5.6 Sol found work that could be precomputed, avoided, or parallelized. With Codex, GPT‑5.6 Sol autonomously rewrote and optimized our production kernels, the core code that executes the mathematical operations that make up the model. This worked in part because we’ve trained GPT‑5.6 to be effective at writing and improving kernels in Triton and Gluon , two open-source GPU programming languages maintained by OpenAI. These efforts, combined with broader kernel advancements from GPT‑5.6 Sol, reduced end-to-end serving costs by 20%. That Luna price drop completely changes the landscape with respect to lower priced models. At $0.20/million tokens for input and $1.20/million for output Luna is now cheaper than Google's Gemini 3.1 Flash-Lite ($.025/$1.50). Anthropic's cheapest current model is Claude Haiku 4.5, and that's $1/$5 - Luna is now 1/5th of that for input, previously it cost the same. My agent.datasette.io demo site was running on Gemini 3.1 Flash-Lite. I've switched it over to Luna. Via Hacker News Tags: ai , openai , generative-ai , llms , anthropic , gemini , llm-pricing
v16.3.0-canary.104
Misc Changes [turbopack] Respect sideEffects in a package.json file with optimizePackageImports : #96383 [turbopack] Keep track of components of the remainder chunk: #96409 Update Rust toolchain to nightly-2026-07-29: #96365 Remove obsolete shouldWaitOnAllReady render option: #96401 Upgrade React from 6cb4322d-20260729 to 0f42eac2-20260730 : #96402 fix: respect htmlLimitedBots in cache components without buffering the full response: #96367 test: expand Cache Components metadata streaming coverage: #96366 fix(create-next-app): make AGENTS.md blue: #95432 [bench skill] Bench Edge builds of React too: #96234 Upgrade React from 1724e9ce-20260729 to 6cb4322d-20260729 : #96389 Fix empty Fragment scroll ownership: #96342 Skip reader task locking for immutable Turbo Tasks reads: #96179 Upgrade React from 96fcba90-20260728 to 1724e9ce-20260729 : #96355 test: reproduce empty parallel route scroll regression: #96323 [ci] Match the test job summary style to the PR test report comment: #95882 [Instant] Point to first blocking await in validation: #96343 Credits Huge thanks to @sampoder , @wbinnssmith , @ztanner , @vercel-release-bot , @styfle , @gaearon , @DavidIlie , @marcoshernanz , @eps1lon , and @lubieowoce for helping!
Investigating three real-world incidents in our cybersecurity evaluations
Investigating three real-world incidents in our cybersecurity evaluations It happened again! This is turning into something of a pattern. Last week OpenAI accidentally exploited Hugging Face when one of their frontier models broke out of a sandboxed container and hacked into Hugging Face to try and get the solutions to the cyber benchmark it was executing. This inspired Anthropic to double-check their own logs, and it turned out they had three similar (albeit less impressive) incidents, the earliest of which played out in April! Of the 141,006 evaluation runs we reviewed, we identified three separate incidents (involving six total runs, four of which impacted the same organization; the other two incidents each happened in independent evaluation runs). [...] In all cases, Anthropic’s evaluation prompt specified to Claude that its environment was a simulation and that it had no internet access. Due to a misunderstanding between us and our evaluation partner, this was not the case, and internet access was available. Because of this, when Claude’s search led it to real systems on the open internet, it treated them as part of the exercise. [...] Operating under the false belief that all accessible entities were intended to be in-scope for the exercise, Claude compromised the impacted organizations’ infrastructure using basic techniques, such as exploiting weak passwords and unauthenticated endpoints. One of the companies was targeted because its name happened to match the fictional name in the eval. The most concerning of the three incidents involved Claude uploading a malware package to PyPI, after a comically convoluted sequence of steps to get an account: [...] in order to create a PyPI account, Claude needed an email address. And in order to create an email address, it needed a phone number. To get a phone number, after failing to find a free phone number service, it tried—and failed—to obtain funds to pay for a phone number through several different means. It finally backtracked, found a free, non-blocked email provider, used this to register a PyPI account, and then used this account to upload malware to PyPI. That package was then installed by a security company that "routinely installs Python packages and scans them for malware", and the executed code was able to exfiltrate credentials back to Claude! Thankfully that package was removed from PyPI by other automated scanners an hour after it was published, but it had still been downloaded and executed on "15 real systems" by that point. It's abundantly clear now that running evals of cyberattack potential in models is a spectacularly risky business. Every AI lab needs to pay attention to this. Keeping a close eye on what's happening in those sandboxes is crucial. Via Hacker News Tags: pypi , python , sandboxing , ai , generative-ai , llms , anthropic , ai-ethics , ai-security-research
AI hedge fund Situational Awareness may have sold its public portfolio, but it still has its Anthropic shares
The former OpenAI researcher’s fund was forced to unwind public equities after leveraged public bets plummeted. But he still has cards to play.
The AI Aesthetic
Article URL: https://blog.jim-nielsen.com/2026/ai-aesthetic/ Comments URL: https://news.ycombinator.com/item?id=49117099 Points: 240 # Comments: 113
Long Term Support Channel Update for ChromeOS
A new LTS version 150.0.7871.213 (Platform Version: 16700.60.0 ), is being rolled out for most ChromeOS devices. If you have devices in the LTC channel, they will be updated to this version. The LTS channel remains on LTS-144 until October 6th, 2026. Release notes for LTC-150 can be found here Want to know more about Long-term Support? Click here Andy Wu Google Chrome OS
Thames Water boss says leakage targets 'not realistic'
Chris Weston said Thames wanted "to do better" but some of its targets were not achievable.
Investors love AI, as long as you’re a cloud host
Amazon isn't slowing down on data center spending — but investors don't seem to mind.
Tim Cook hints at iCloud Plus tier for AI power users
Apple may allow users to pay to increase their AI usage limits. During an earnings call on Thursday, Apple CEO Tim Cook said that he believes people will want to use Apple Intelligence and the upcoming Siri AI "a lot," adding that "we will have some kind of upgrade possibilities on iCloud Plus where people […]
CISA Urges Water Sector to Protect OT After Coordinated Attacks on PLCs
CISA is urging water and wastewater utilities to lock down internet-exposed controllers, days after intrusions hit dozens of Minnesota systems. The post CISA Urges Water Sector to Protect OT After Coordinated Attacks on PLCs appeared first on SecurityWeek .
Xbox CEO lays out priorities in memo after major ‘reset’
After a massive Xbox "reset" that laid off thousands of employees and spun off four studios, Xbox CEO Asha Sharma wants to get Xbox back to growth. In a memo obtained by The Verge, Sharma told staff by the end of fiscal year 2027 (which runs through next June), "we will return XBOX to player […]
Are Iran and Ukraine wars getting tangled in Caspian Sea?
Tehran has threatened a response after a Ukrainian strike on an Iranian cargo ship in the Caspian Sea that Kyiv says was carrying weapons to Russia. Still, Ukraine says it wants to avoid escalation with Tehran.
Spain will send troops to Ceuta to help with border security
Authorities in the Spanish city of Ceuta asked Madrid to declare a national emergency and send military assistance. Crowds of people breached the border of the territory next to Morocco as police lost control.
Hundreds of migrants swim from Morocco to Spanish enclave of Ceuta
The territory's leader said the influx is overwhelming resources and urged the Spanish government to intervene.
Coinbase shares fall after crypto exchange posts disappointing second-quarter results
Coinbase's Thursday report marked its third straight quarter missing Wall Street's forecasts for revenue and earnings.
Bank of America to Acquire Cybersecurity Firm MDSec
The acquisition will add approximately 65 cybersecurity professionals to Bank of America’s operations in the United Kingdom. The post Bank of America to Acquire Cybersecurity Firm MDSec appeared first on SecurityWeek .
Will Saudi Arabia be dragged further into the war in Iran?
Anger in Iraq after US-Saudi strikes on paramilitary group with ties to Tehran.
Prague Zoo keeps polar bears cool with ice amid heatwave
Keepers at Prague Zoo used tonnes of ice to provide much relief to animals as Europe grappled with scorching heat.