Author: Anthony John Padilla

{ "social": { "email": "content.reviewer90@ext.softonic.com", "facebook": "", "twitter": "", "linkedin": "" }, "ja-JP": "", "de-DE": "Anthony Padilla ist ein auf den Philippinen ansässiger Schriftsteller mit Erfahrung in redaktionellen, digitalen und markenorientierten Inhalten. Seine Arbeit konzentriert sich darauf, komplexe Ideen in klare, verständliche Texte umzuwandeln.", "en-US": "Anthony Padilla is a Philippines-based writer with experience across editorial, digital, and brand-focused content. His work centers on translating complex ideas into clear, accessible copy, with a strong emphasis on structure, accuracy, and user-focused communication. He has contributed to long-form articles, product and platform copy, and marketing-led editorial pieces.", "es-ES": "Anthony Padilla es un escritor con base en Filipinas con experiencia en contenido editorial, digital y enfocado en marca. Su trabajo se centra en traducir ideas complejas en textos claros y accesibles.", "fr-FR": "Anthony Padilla est un écrivain basé aux Philippines ayant de l'expérience dans le contenu éditorial, numérique et axé sur la marque. Son travail consiste à traduire des idées complexes en textes clairs et accessibles.", "it-IT": "Anthony Padilla è uno scrittore con base nelle Filippine con esperienza in contenuti editoriali, digitali e focalizzati sul brand. Il suo lavoro si concentra sulla traduzione di idee complesse in testi chiari e accessibili.", "nl-NL": "Anthony Padilla is een schrijver gevestigd op de Filipijnen met ervaring in redactionele, digitale en merkgerichte content. Zijn werk richt zich op het vertalen van complexe ideeën naar helder, begrijpelijke teksten.", "pl-PL": "Anthony Padilla to pisarz mieszkający na Filipinach z doświadczeniem w tworzeniu treści redakcyjnych, cyfrowych i skoncentrowanych na marce. Jego prace skupiają się na przekształcaniu skomplikowanych idei w jasne, zrozumiałe teksty.", "pt-BR": "Anthony Padilla é um escritor baseado nas Filipinas com experiência em conteúdo editorial, digital e focado em marca. Seu trabalho se concentra em traduzir ideias complexas em textos claros e acessíveis." }

Claude Mythos flags 23,019 software flaws: 21,000 still await human review

By June 2026, Anthropic said its AI system, Claude Mythos, had produced 23,019 vulnerability reports across open-source software. Only about 1,900 had been reviewed by outside parties. That meant more than 21,000 findings were still unverified. Even so, Anthropic says 1,726 of the reviewed reports turned out to be real vulnerabilities, which helps explain why AI-driven bug hunting can feel like a real shift in security work. But people still have to do the hard part, because severity was often rated too high and the choke point is moving away from discovery and toward validation and patching.

That need for human review shows up pretty clearly in the severity calls. Across 27 Common Vulnerabilities and Exposures, or CVE, cases, Anthropic’s Claude Mythos labeled eight as critical. After review, only one of those still counted as critical, and 13 of the 27 original ratings were judged too high. Temporal Server is a good example. A flaw there was first tagged as critical, then later dropped to low because exploiting it depended on rare setups, unusual deployment conditions, or privileged access. At that point, maintainers still had to sort out exploitability, business impact, and patch priority.

Anyone who follows software security should keep an eye on this, because the bottleneck is shifting from finding bugs to checking them and fixing them, and monthly CVE disclosures rose 145% in the two years leading up to June 2026. These reports hit open-source projects, not some consumer app download page, and researchers are divided. On one side, cheaper and faster auditing. On the other, the risk that machine-generated disclosures will swamp teams, maintainers, and bug bounty programs as the number of discovered bugs keeps climbing.

ChatGPT, Claude, and Grok all go down at once: widespread outages hit

OpenAI’s ChatGPT, Anthropic’s Claude, and xAI’s Grok all went down on September 3, 2026, leaving people unable to load chats, send prompts, or get answers back.

ChatGPT Download

The flood of outage reports hit between 10:30 AM and 11:00 AM ET, based on Downdetector data. At one point, Downdetector showed more than 35,000 reports in the U.S. for ChatGPT, 1,400 for Claude, and 1,200 for Grok. In India, the counts were 2,800, 170, and 40. OpenAI posted on its status page that it was seeing “elevated errors across ChatGPT and Codex,” later said it had put in a fix, and kept watching recovery. Anthropic’s status page first named Mythos/Fable 5.1, Mythos/Fable 5, Opus 5, Opus 4.8, and Opus 4.6, then later narrowed the issue to Opus 4.8 and Opus 5. Grok users reported trouble on both the web version and the mobile app, and a smaller number of users said Google Gemini was also acting up.

For anyone using these tools for coding, writing, research, or customer support, outages like this can bring work to a halt pretty quickly, especially in a market where so much depends on a small handful of platforms. Some people speculated that shared cloud infrastructure might be involved, including possible Azure ties, but none of the companies confirmed a single cause. If you want to follow the recovery, check the status pages for OpenAI, Anthropic, and xAI.

AWS, SANS, OWASP and NIST warn: system prompts aren’t a security boundary

Amazon Web Services (AWS), the SANS Institute, the Open Worldwide Application Security Project (OWASP), and the U.S. National Institute of Standards and Technology (NIST) are now saying that if you’re deploying AI agents, you can’t treat system prompts as a real security boundary. That shift puts prompt injection and runtime controls back at the center of the conversation.

OWASP still lists prompt injection among the top large language model (LLM) risks in its updated Top 10, because attacker-controlled content can steer models into exposing sensitive files or taking actions they shouldn’t. NIST’s 2026 Agent Standards Initiative came after research that, according to NIST, found task hijacks worked 81% of the time. OWASP has also added an Agent Control Standard.

If you’re running agents across internal tools, support, development, or operations, the warning from AWS, SANS, OWASP, and NIST is pretty direct. Put controls in place when an agent reads data, calls tools, or executes actions. Give the agent the same permissions as the user, or fewer. Start from deny, not allow.

You can see the gap in recent adoption data. Deployed agents doubled in four months. Thirty-eight percent of organizations are running more than 100, and more than 80% say they could stop unauthorized data access. Even so, nearly 9 in 10 still reported an agent-related incident, including a 2026 OpenAI plugin ecosystem supply-chain attack that affected 47 enterprises, along with compromises in multi-agent infrastructure.

The latest guidance and standards from AWS, SANS, OWASP, and NIST are available in their published reports and security documentation.

Canadarm is now the 300th IEEE Milestone for 2026

The original Canadarm has picked up another honor. IEEE, the Institute of Electrical and Electronics Engineers, has named the Canadian-built robotic arm from the Space Shuttle its 300th Milestone for 2026, recognizing a technology that marked a major technical achievement and left lasting benefits for society.

NASA Download

It also puts the spotlight back on the role Canadarm played in NASA’s Space Shuttle program. According to NASA, the arm flew on 90 of the shuttle’s 135 missions, where it was used to deploy, retrieve, and repair spacecraft and other payloads in orbit.

Canadarm first flew in 1981. NASA describes it as the world’s first spaceborne robotic system. At 15.2 meters long and 410 kilograms, it could handle payloads of up to 266,000 kilograms in microgravity, and do it with remarkable precision.

That turned Canadarm into one of the signature tools of the shuttle era. It made delicate, difficult, risky, and sometimes flat-out impossible jobs possible, including major work on the Hubble Space Telescope. NASA and IEEE both point to that record when they describe how the arm helped turn the shuttle into a true orbital work platform.

For anyone who follows spaceflight history, this is a fitting piece of recognition. Canadarm was built by SPAR Aerospace, now part of MDA Space, for about $108 million, according to IEEE. Over time it became a Canadian engineering icon and a cornerstone of Canada’s partnership with NASA. Few pieces of space hardware have had a run this long, or this influential.

That legacy is still easy to trace. You can see it in Canadarm2 and Dextre aboard the International Space Station, in Canadarm3 for Artemis lunar missions, and in the milestone coverage from IEEE and NASA. So far, most of the reaction has come from North America, much more than from the European Space Agency, the Japan Aerospace Exploration Agency, or Roscosmos.

Stellarium Download

Askeal launches source-backed cybersecurity assistant for SOC teams, now in beta

Askeal has opened beta access to its cybersecurity assistant for security operations center, or SOC, teams and IT professionals. The pitch is straightforward: fast answers in plain language, backed by visible citations, instead of opaque black-box output, the company says.

Wireshark Download

Askeal says those answers are built from vetted community sources, including security vendors, researchers, and frontline practitioners. And when reliable sources disagree, the tool shows both views instead of flattening them into a single answer.

Users can also send in questions with files or other supporting evidence, according to Askeal. For analysts working through 50 or more alerts a day, that could mean less manual review. The company positions the tool alongside endpoint detection and response, or EDR, and extended detection and response, or XDR, not as a replacement. If you work in a SOC, that’s the part that stands out: you can look at the evidence yourself before you act.

Askeal was co-founded by CEO Roxane Suau in August 2025 and has raised $1.1 million in pre-seed funding to help with market entry and international expansion, the company says. After launching the beta in February 2026, Askeal says it passed its goal of 300 testers, reaching 500 users across 69 countries in roughly two and a half months. About half were SOC analysts.

In June 2026, the company added SOC Agent for broader investigations. Askeal says it handles data ingestion, indicator extraction, evidence correlation, and recommended next steps. The beta is available to try now.

Chatbots may be making conversation more robotic: a new study warns

A recent paper on chatbots makes a simple point that’s hard to shake: the more human chatbots sound, the more people may start reshaping their own language around them.

The paper gives that shift a name, “robotoid humanness.” It describes a two-way process where automated systems get better at sounding human, while people slide the other way and start sounding more machine-like in their pacing, prompts, and overall interaction style. You can already see the pattern most clearly in customer service, where chat assistants tend to reward concise, structured messages that are concise, structured, and purely transactional.

ChatGPT Download

One Harvard Business School study found that automated help let agents respond 22% faster and pushed customer sentiment up by 0.45 points on a five-point scale. But there’s a catch. Replies that are too fast and too polished can chip away at trust and satisfaction once people start thinking they’re talking to a bot, which creates an uncanny effect for companies trying to automate support.

The paper’s bigger concern is what happens over time. Heavy chatbot use could feed social deskilling, weaken critical-thinking habits, erode language, and even encourage emotional dependency. It also admits the evidence is still narrow: most of the research comes from Western, English-speaking settings, with much less data on how culture, class, language, and life in developing or rural areas might change the picture.

If you use chat support a lot, this is one to keep an eye on as chatbots become the default layer between people and businesses.

Claude Code weekly limits updated: less for most users

Anthropic has changed Claude Code’s weekly limits. The company says the new permanent policy sits about 25% above the original baseline from September 14, but it’s still roughly 17% below the temporarily boosted allowance users have access to today.

On paper, that sounds like an increase. In day-to-day use, it’s a cut from what people have been getting. If your old cap was 100 units, the temporary 50% boost Anthropic introduced on May 13, then extended four times, pushed that up to 150. Under the new permanent policy, you’d get 125. Anthropic later clarified that this is about 17% less than what you can use today. The change affects Anthropic’s Pro, Max, Team, and seat-based Enterprise plans, and the company says it’s meant to keep the platform sustainable while giving users better usage controls and clearer visibility into those limits.

Users pushed back because Anthropic’s first announcement highlighted the permanent increase without plainly saying that the current boosted allowance was going away. That led to an X Community Note calling the move a downgrade. The frustration is easy to understand when the summer boost had, for a lot of developers, stopped feeling temporary and started feeling like the default plan.

If you use Claude Code for long debugging sessions, digging through large codebases, or other work that burns through a lot of context, this feels more like a downgrade than an upgrade. Even more so because opaque quotas already make serious planning harder than it should be. By Anthropic’s own figures from November 2025, coding is one of the biggest use cases for Claude models, and software error fixing alone accounts for 6% of all usage. It also lines up with the broader pullback happening across GPU-heavy coding tools.

Claude Code is still available through Anthropic’s Pro, Max, Team, and seat-based Enterprise plans.

Aurora ransomware gang used Cursor to speed up attacks: 30% to 50% faster intrusions

From April through July 2026, Aurora ransomware operators used Cursor, the coding assistant recently acquired by SpaceX, during intrusions inside victim networks, according to new research from Gambit Security and CloudSEK.

The researchers say Cursor wasn’t used to build the ransomware itself. Once Aurora was already inside, the operators leaned on it for reconnaissance, exploitation, credential theft, network scanning, command generation, lateral movement, and command-line troubleshooting when something broke.

Gambit Security and CloudSEK say that turned Cursor into an always-available operations helper, something attackers could query for the next move after a tool failed or a command didn’t work. They estimate it cut intrusion time by 30% to 50% and reduced the need for specialist skills.

If you track cybercrime or spend your time defending enterprise networks, this report is worth reading.

According to Gambit Security and CloudSEK, data from an exposed Aurora affiliate server showed activity from April to July 2026 targeting more than 20 organizations in at least nine countries, including the US, Belgium, Germany, Scotland, Italy, and Argentina. The researchers say the data suggests Cursor was a routine part of the workflow, not some side experiment.

They also say Aurora operators got around some guardrails by framing malicious requests as penetration-testing or simulation exercises. That raises the stakes for vendors, who need stronger abuse checks, and for defenders, who may need more automated defenses to keep up.

You can read the full Gambit Security and CloudSEK findings online.

FBI cripples China-linked hacking network: NASA, DOJ and Senate among targets

The FBI says it has disrupted the QScan and QTRouter network, a system authorities say had been used since at least 2018 to go after NASA, the US Department of Justice, the Federal Reserve, the US Senate, the US National Institutes of Health, and other American agencies and pieces of infrastructure.

Court filings say FBI agents seized malware-linked domains to cut off attacker communications and knock both platforms offline. Investigators connect them to QTFY, which the US Department of Justice identifies as part of Nanjing Xinjiuwei Network Technology Company. US officials say that company is tied to China’s Ministry of State Security and the People’s Liberation Army. Beijing denies those claims.

Officials say QScan was used to find and compromise exposed IoT devices, especially small office and home office routers and cameras. QTRouter then used those hacked devices as relays across more than 130 countries, making the traffic look local, or at least not like it was coming from China.

If you’ve got smart devices at home or at work, pay attention to this. Change default passwords, install firmware updates, and replace hardware that no longer receives patches. the warning comes as the US keeps moving against Volt Typhoon, Flax Typhoon, and Mustang Panda.

You can read the court documents, along with statements from the FBI and the US Department of Justice. US officials say the takedown hits an operation that helped cover up years of intrusions into American targets.

GPUThor disclosed: it bypasses ECC on NVIDIA RTX workstation GPUs

Researchers disclosed GPUThor today, a GPU-based Rowhammer attack that targets GDDR6 memory on NVIDIA Ampere workstation GPUs. In their write-up, repeatedly hammering that memory caused bit flips that led to corruption, crashes, and, in one demonstration on an NVIDIA RTX A6000, a full compromise of the host.

They say GPUThor works on RTX A4000 , RTX A4500, RTX A5000, and RTX A6000 cards. It can also push error-correcting code, or ECC, past what it can reliably handle by forcing multi-bit errors, cases that may be detected but not corrected, or even silently corrected the wrong way. In the researchers’ testing, GPUThor was roughly 4,548 to 23,597 times more effective than the earlier GPUHammer technique at producing exploitable flips directly on the GPU.

If you’re running shared cloud, research, or high-performance computing GPU workloads, this deserves real attention. In the team’s tests, an unprivileged program was able to corrupt GPU page tables, turn GPU memory corruption into root-level access on the host, and also trigger GPU resets about every two hours. In some cases, it eventually went far enough to raise hardware-replacement flags.

The researchers say the findings affect NVIDIA GDDR6 workstation setups. NVIDIA recommends ECC along with input-output memory management unit, or IOMMU, and direct memory access, or DMA, isolation. The paper argues that operators may need more than that, including stronger multi-bit ECC, DRAM-level mitigations, close monitoring of ECC error logs, and tighter controls on untrusted GPU workloads.