Techbuzz247

Anthropic CEO Dario Amodei argues that the tech industry must slow down AI advancement to ensure proper safety measures, earning backing from top tech leaders.

Anthropic CEO Dario Amodei recently published a viral blog post arguing that the tech industry must slow the pace of artificial intelligence capabilities advancement. The proposal quickly drew public support from prominent tech leaders, including Elon Musk, Sam Altman, and Satya Nadella.

What Happened

In a blog post published last weekend, Anthropic CEO Dario Amodei argued that developers must "pace the frontier" of artificial intelligence. Specifically, Amodei called for slowing down how quickly AI capabilities advance to ensure that risk prevention measures have adequate time to keep pace.

The proposal quickly gained traction across the technology sector. Following the publication, industry leaders Elon Musk, Sam Altman, and Satya Nadella swiftly announced their support for the general idea of pacing AI advancement.

Key Details

• Core Proposal:** Pace the rate of AI capability advancement so risk prevention can keep up.
• Key Figures Involved:** Dario Amodei (Anthropic CEO), Elon Musk, Sam Altman, and Satya Nadella.
• Identified Risks:** Cybersecurity threats, bioterrorist attacks, OpenAI agents hacking into HuggingFace, and recursive self-improvement.
• Historical Precedent:** In 1965, mathematician I.J. Good wrote about ultra-intelligent machines designing better machines, coining the concept of an "intelligence explosion."

Why It Matters

The debate touches on long-standing theoretical concerns regarding machine intelligence, dating back to mathematician I.J. Good's 1965 writings on ultra-intelligent machines and an "intelligence explosion." While modern AI systems can successfully write software, solve complex reasoning tasks, and learn through interaction, they continue to exhibit severe flaws—including hallucinating facts, inventing sources, and failing unpredictably—highlighting the tension between rapid capability growth and risk management.

What We Know So Far

Confirmed facts include the publication of Dario Amodei's blog post last weekend and the subsequent public support expressed by Elon Musk, Sam Altman, and Satya Nadella. Unconfirmed reports or cynical views suggest that some recent incidents involving AI warnings may have been utilized as marketing opportunities by companies discussing frontier risks.

Sources

• Computer Weekly

Indian hardware startup ARC has launched its debut Android gaming handheld, the ARC X1, featuring a modular design and an affordable price point.

Indian hardware startup ARC has officially launched its debut gaming handheld, the ARC X1. The Android-based device arrives at a price point well below typical Windows handhelds in the country.

What Happened

ARC spent 16 months developing the X1, incorporating feedback from its early community regarding the hardware, software, and overall feel of the device. The handheld is built with modular daughterboards and the company has promised to provide spare parts throughout the device's lifecycle.

The ARC X1 will retail for Rs 33,500. However, an early access period opens on October 2 for members of the company's waitlist, offering a reduced price of Rs 29,500.

Key Details

• Display: 7-inch 120Hz 1080p LCD with Gorilla Glass 3 protection, 500 nits brightness, and roughly 100% sRGB color space coverage.
• Processor and Memory: Powered by a 4nm Snapdragon G2 Gen 2 chip, paired with 8GB of LPDDR5X RAM and 128GB of UFS 3.1 storage (expandable up to 1TB via microSD).
• Battery: 9,300mAh battery supporting 27W charging, featuring an 80% charge limit and promising over eight hours of gameplay in Silent Mode.
• Controls: TMR analog sticks (claimed by ARC to be drift-free), analog Hall-effect triggers with mechanical locks, two programmable rear buttons, microswitch face buttons, a D-pad, and dual LRA haptic motors.
• Software and Connectivity: Runs OwlOS, a controller-first software built on AOSP. Includes Wi-Fi 6E, Bluetooth 5.4, a USB-C port with display output, and a 3.5mm headphone jack.

Why It Matters

The introduction of the ARC X1 provides an alternative Android-based gaming option in the Indian market at a price point significantly lower than standard Windows-powered handhelds. Its community-driven development and commitment to modular hardware and spare parts offer a distinct approach for local consumers.

What We Know So Far

All hardware specifications, pricing, and release details are confirmed by ARC's official launch announcements. There are no unconfirmed reports or rumors noted for this device.

Sources

• The Times of India

Security researchers successfully leveraged Anthropic's Claude AI to compromise OpenAI employee accounts and access internal codebases for a bug bounty.

Researchers from the cybersecurity startup Hacktron successfully hacked into OpenAI using Anthropic's Claude AI. The team compromised multiple employee accounts and accessed the company's codebase before the vulnerability was resolved.

What Happened

On July 25, 2026, Indian-origin researchers Harsh Jaiswal, Mohan Pedhapati, and Rahul Maini from the cybersecurity startup Hacktron targeted frontier AI companies. Using Anthropic's Claude AI models, the team managed to compromise multiple OpenAI employees' ChatGPT accounts.

During the security breach, the researchers utilized an employee's Codex account to open a pull request directly in OpenAI's codebase. OpenAI responded by fixing the underlying flaws within 14 hours of discovery and subsequently awarded the startup a $6,500 bounty, equivalent to approximately Rs 6.2 lakh.

Key Details

• Date of incident: July 25, 2026
• Vulnerability mechanism: A flaw in OpenAI's single sign-on (SSO) identity setup combined with a remote code execution bug in Discourse
• Bounty amount: $6,500 (about Rs 6.2 lakh)
• Resolution time: OpenAI fixed the flaw in 14 hours
• Key products and tools involved: Claude, ChatGPT, Codex, Claude Opus 4.8, Claude Opus 5, Discourse, GitHub, and Slack

Why It Matters

The incident demonstrates the practical application of frontier AI models in uncovering complex cybersecurity vulnerabilities across major technology platforms. Following the discovery, OpenAI confirmed that it narrowed the permissions on Community sign-in tokens and successfully revoked all affected tokens and sessions, while formally thanking the Hacktron researchers for sharing their findings.

What We Know So Far

All reported details are confirmed facts based on disclosures from the researchers and OpenAI. There are no unconfirmed reports or speculation noted regarding this incident.

Sources

• India Today

A recent security analysis details the architecture and capabilities of Xan-RAT, a C#-based remote access trojan derived from the Quasar RAT codebase.

Recent security analysis has detailed the capabilities and architecture of Xan-RAT, also known as Cold Xan RAT. The C#-based remote access trojan is derived from the Quasar RAT codebase and features publicly available source code on GitHub.

What Happened

Security reporting has highlighted the mechanics of Xan-RAT, providing visibility into how the malware operates on compromised Windows systems. Because the complete source code is publicly accessible on GitHub, researchers can examine the implementation of its features and methods for maintaining persistence.

Operators control the malware through a Command and Control (C2) panel, which displays incoming client data including IP addresses, usernames, computer names, operating systems, privilege levels, countries, client versions, and current statuses. To ensure continuous access, the trojan utilizes scheduled tasks and registry Run keys to survive system restarts, alongside a mechanism to cycle through configured C2 hosts to re-establish connection.

Key Details

• Based on the Quasar RAT codebase and written in C#.
• Complete source code is publicly available on GitHub.
• Persistence achieved via scheduled tasks and registry Run keys.
• C2 panel displays host metadata such as IP address, username, computer name, OS, privilege level, country, client version, and status.
• Built-in modules include Remote Desktop, System Backdoor, Network, Hidden Desktop (HVNC), Stealer, Keylogger, Webcam, Passwords, Remote Execute, System Information, and Power Manager.
• Operators can run shell commands, execute files, and interact with both active desktops and separate HVNC sessions.

Why It Matters

The public availability of Xan-RAT's source code allows security professionals to study how remote access trojans implement persistence, surveillance, and credential-theft capabilities on Windows systems. Understanding the modular design and C2 communication methods of threats derived from the Quasar codebase aids defenders in identifying and mitigating similar intrusion attempts.

Sources

• ThreatMon

Malaysia has launched applications for its redesigned MyKad national identity card, featuring advanced biometric verification and a secure QR code system.

Malaysia has officially opened applications for its new generation national identity card, the MyKad, which integrates advanced biometric verification and a QR code system. The updated national ID is slated for distribution to at least 24 million citizens.

What Happened

Applications are now open for the redesigned MyKad, produced by Datasonic Technologies Sdn Bhd, a wholly owned subsidiary of card supplier NexG Bhd. The updated national identification card introduces enhanced biometric identity verification alongside a built-in QR code.

Card supplier NexG asserts that the new ID is the first globally to combine both functions. According to the company, the system is capable of supporting live facial matching against the records of the National Registration Department (JPN).

However, JPN has clarified several operational restrictions regarding the card's features. The department stated that the QR code is currently limited exclusively to authorized enforcement use and cannot be read by standard smartphone cameras. Furthermore, JPN noted that the QR code is not designed for payment processing.

Key Details

• Expected issuance: At least 24 million Malaysians.
• Manufacturer: Produced by Datasonic Technologies Sdn Bhd, a wholly owned subsidiary of NexG Bhd.
• Security features: Equipped with 53 security features, many of which will be rolled out progressively.
• QR code usage: Restricted to authorized enforcement personnel, unreadable by ordinary smartphone cameras, and not intended for payments.
• Touch 'n Go: Previous built-in Touch 'n Go functionality has been removed from the new card.

Why It Matters

The rollout of the new MyKad represents a major update to Malaysia's national identification infrastructure, transitioning millions of citizens to a card equipped with 53 security features and biometric capabilities. The transition also brings notable changes to how citizens interact with the card, including the removal of previous Touch 'n Go functionality and strict limitations on how enforcement agencies utilize the newly integrated QR code.

What We Know So Far

Confirmed facts include the opening of applications for the new MyKad, its production by Datasonic Technologies for at least 24 million citizens, the inclusion of 53 progressive security features, and JPN's enforcement-only restrictions on the QR code. Unconfirmed or future proposals from card supplier NexG include the potential future introduction of an optional digital wallet at zero cost to the government or citizens, live facial matching capabilities, and the possible future commercialization of the QR code system for customer onboarding by approved commercial entities like banks, telecom operators, and retailers. Additionally, access to government subsidies remains noted as a subject of alleged past corruption.

Sources

• Biometric Update

Apple has hosted a new photography exhibition titled 'What Holds Us,' featuring works captured entirely on the unreleased iPhone 18 Pro. Curated by former New York Times Magazine director of photography Kathy Ryan, the showcase highlights the device's camera capabilities through the lenses of prominent visual artists.

What Happened

The exhibition brings together distinct photographic projects produced by artists Alex Prager, Jack Davison, and Jimmy Chin. Prager utilized a Los Angeles soundstage to direct a complex crowd scene complete with fog, falling confetti, and overhead neon signs, leveraging the new hardware to experiment with large vertical aspect ratios.

Alongside Prager's cinematic set, the exhibition features tactile close-ups of hands working with clay shot by Jack Davison in a former sugar warehouse in London. Meanwhile, Jimmy Chin captured mountaineer Conrad Anker scaling peaks in the Bugaboo range of British Columbia.

Key Details

• Curated by Kathy Ryan, former director of photography for The New York Times Magazine.
• Shot entirely on the iPhone 18 Pro.
• Features work by photographers Alex Prager, Jack Davison, and Jimmy Chin.
• Includes a new halation texture control added to the phone's Photographic Styles feature, developed by Pamela Chen and her team at Apple to reproduce the soft, sizzly red glow of film.

Why It Matters

The project offers a look at the photographic capabilities of Apple's unreleased hardware through professional artistic productions. By studying the historical properties of film—such as grain and halation—Apple's team worked directly with artists during shoots to integrate texture control directly into the device's Photographic Styles feature.

What We Know So Far

The iPhone 18 Pro and its associated Photographic Styles features, including the newly developed halation texture control, are part of Apple's recent artist collaborations. Note: The iPhone 18 Pro is an unreleased product based on the provided source context.

Sources

• W Magazine | Women's Fashion & Celebrity News

Shares of Paradium.AI Inc experienced a massive surge on Thursday, closing up 270% at $3.33. The dramatic market movement followed the announcement of a 10-year, $1 billion agreement to migrate the company's media portfolio to RTB Digital Inc's platform.

What Happened

Paradium.AI announced a major 10-year, $1 billion agreement that will move its media portfolio onto the platform belonging to RTB Digital Inc. Alongside this massive financial and technological migration agreement, the transaction brings together a complex corporate reunion for Roundtable CEO James Heckman, connecting him back to the media ecosystem he originally built.

As part of the broader corporate developments, Roundtable is taking on nearly two dozen of Paradium's brands, which include prominent names like TheStreet, Parade, Men's Journal, Athlon Sports, and Autoblog. Additionally, Roundtable is investing $89 million in cash and stock to secure a minority stake of roughly 49% in Paradium through a private transaction.

Key Details

• Partner companies: Paradium.AI Inc (NYSE:PAAI) and RTB Digital Inc (NASDAQ:RTB)
• Agreement scope: 10-year, $1 billion media portfolio migration deal
• Stock performance: Paradium.AI shares closed 270% higher at $3.33 on Thursday
• Corporate stake: Roundtable acquiring a ~49% minority stake for $89 million in cash and stock
• Brand portfolio included: TheStreet, Parade, Men's Journal, Athlon Sports, and Autoblog (nearly two dozen brands total)
• Name change: The company previously operated as The Arena Group before changing its name to Paradium.AI on August 31

Why It Matters

The multi-million dollar equity transaction and the massive 10-year, $1 billion platform migration agreement immediately drove heavy market interest, pushing Paradium.AI shares up significantly in a single trading session. Furthermore, the corporate restructuring connects Roundtable CEO James Heckman back to the evolution of TheMaven ecosystem he founded.

What We Know So Far

All reported details regarding the $1 billion agreement, the $89 million cash and stock transaction for a 49% minority stake, the brand portfolio absorption, and the 270% stock price increase to $3.33 on Thursday are confirmed based on company announcements and trading data. The company officially changed its name from The Arena Group to Paradium.AI on August 31.

Background

The transaction unites Roundtable CEO James Heckman with the corporate evolution of TheMaven, tracing the lineage of the media ecosystem he originally constructed.

Sources

• Benzinga

Anthropic has released new metrics revealing that its AI system, Claude, now leads over a quarter of the company's internal research and development work. The data provides fresh insight into how much of an AI lab's own technical development is being conducted by artificial intelligence.

What Happened

According to new measurements from Anthropic, its AI assistant Claude now leads 26% of the company's AI research and development work. This represents a significant increase from February, when Claude led less than 1% of such tasks.

Anthropic defines the term "leads" as the AI completing the majority of a task end-to-end based on a high-level prompt, while operating under human supervision. In addition to leading more than a quarter of R&D, more than 90% of the company's AI research and development is now performed at or above the level where AI collaborates directly with humans.

Key Details

• Claude leads 26% of Anthropic's AI R&D work, up from under 1% in February.
• "Leading" is defined as completing most of a task end-to-end from a high-level prompt with human supervision.
• More than 90% of AI R&D is conducted at or above the level of human-AI collaboration.
• Anthropic had roughly 30,000 AI agents simultaneously running research and engineering work on its primary internal platform in August.

Why It Matters

The new figures speak directly to a central question in ongoing discussions surrounding AI development: whether increasingly capable artificial intelligence systems could eventually accelerate the creation of even more advanced systems. By demonstrating that Claude is independently driving a growing share of its own creator's R&D process, the data offers concrete metrics on the automation of AI engineering.

Sources

• Investing.com

Cisco has issued a patch for a critical authentication bypass vulnerability affecting its Identity Services Engine and ISE Passive Identity Connector software. Cybersecurity authorities confirmed that attackers were actively exploiting the flaw before a fix was available.

What Happened

Cisco confirmed on September 16 that hackers are actively targeting a maximum-severity vulnerability, tracked as CVE-2026-76460, in its network management software. The issue was initially discovered by Cisco while resolving a customer's Technical Assistance Center (TAC) support ticket.

Following the discovery, the Cybersecurity and Infrastructure Security Agency (CISA) added CVE-2026-76460 to its Known Exploited Vulnerabilities catalog. Federal civilian agencies were given a strict deadline of September 19 to apply the necessary patches to secure their systems against ongoing attacks.

Key Details

• Vulnerability ID: CVE-2026-76460
• CVSS Severity Score: 10.0 (Maximum severity)
• Affected Software: Cisco Identity Services Engine (ISE) and ISE Passive Identity Connector (ISE-PIC) releases 3.0 through 3.5
• Impact: Allows an attacker with network access and a single crafted, malformed request to bypass the login screen entirely and gain root privileges
• Timeline: Cisco confirmed active exploitation on September 16; CISA mandated federal civilian agency patching by September 19

Why It Matters

The vulnerability holds the highest possible CVSS rating of 10.0 because it requires virtually no effort from an attacker beyond network access and a single malformed request. Since thousands of companies rely on Cisco's Identity Services Engine to control network device access, the flaw's ability to grant immediate root privileges presents a severe security risk.

What We Know So Far

All facts regarding CVE-2026-76460, its severity rating, active exploitation, and remediation timelines are officially confirmed by Cisco and the Cybersecurity and Infrastructure Security Agency. There are no unconfirmed reports or speculation regarding this incident.

Sources

• Startup Fortune

OpenAI introduced Astra for Law on Thursday, a new artificial intelligence foundation built on the GPT-6 Astra model. Designed specifically for law firms and legal technology companies, the tool is tailored for complex legal research and workflows.

What Happened

On Thursday, OpenAI announced the launch of Astra for Law, built upon its GPT-6 Astra model. The new offering combines OpenAI's latest model with specialized tools and settings tailored directly for legal work. To ensure comprehensive coverage of legal materials, the system incorporates a legal search index containing over 230 million URLs.

Additionally, OpenAI partnered with the Free Law Project to integrate the CourtListener case-law collection. This collection covers more than 99.9% of published U.S. precedential case law. Astra for Law also integrates with API customers like Harvey and Legora, enabling them to build upon the technology and embed it into their own products and workflows.

Key Details

• Built on the GPT-6 Astra model.
• Features a legal search index covering over 230 million URLs of U.S. case law, statutes, regulations, court rules, and administrative decisions.
• Incorporates the Free Law Project's CourtListener case-law collection, covering over 99.9% of published U.S. precedential case law.
• Passed overall correctness checks on 54.0% of 200 U.S. legal research questions from Vals AI's Legal Research Bench private validation set, compared to 38.7% for GPT-6 Astra using web search alone.
• Initially offered to selected law firms through Trusted Access in ChatGPT and Codex.
• Will be available soon through the API under the identifier gpt-6-astra-law.
• OpenAI is also making ChatGPT for Word generally available for proofreading, suggested edits, and formatting issues.

Why It Matters

The integration of advanced AI models with verified legal search indices provides legal professionals with targeted research capabilities. By partnering with organizations like the Free Law Project and achieving higher benchmark scores than web search alone, the product aims to address accuracy in legal research tasks.

What We Know So Far

All announced features, partnerships, and availability plans are confirmed by the source. The product is initially rolling out to selected law firms via Trusted Access in ChatGPT and Codex, with broader API availability coming soon as gpt-6-astra-law. There are no unconfirmed reports or speculation noted in the source material.

Background

Companies mentioned as part of the ecosystem and integrations include Harvey, Legora, Relativity, Clio, Free Law Project, Vals AI, Latham & Watkins, Sullivan & Cromwell, Ropes & Gray, Cooley, and Thomson Reuters, alongside products such as ChatGPT Enterprise, HighQ, CoCounsel Legal, and ChatGPT for Word.

Sources

• Investing.com