Catch up on The Next Web articles in minutes

ShinyHunters claims to have defaced Cl0p's dark web leak site, citing an unauthenticated file-upload flaw in Grav. They assert possession of Cl0p's source code and private keys, threatening further extortion. Both groups have a history of significant cyberattacks.

SoftBank is issuing approximately $11 billion in bonds to fund its next payment to OpenAI. This bond sale is a refinancing measure to replace bridge loans used for the initial investment. The pricing of these bonds will be a key indicator of SoftBank's financial standing.

Firmus is reportedly set to launch its $5bn ASX IPO on October 6th, with trading to commence on October 22nd. The company has secured significant contracts, including with OpenAI, and has attracted major investors like Nvidia and Blackstone. However, details of the IPO timetable are not yet public, and expert concerns about escrow arrangements have been raised.

Insta360's significant investment in memory chips contrasts with GoPro's sale, highlighting industry-wide supply chain pressures. While Insta360 navigates these challenges, GoPro struggles, and DJI faces regulatory hurdles. Insta360's market growth and strategic expansion suggest a strong position in the evolving camera industry.

Z.ai's ZCode uploaded encrypted developer workspaces to Alibaba Cloud by default, with the private key held by Z.ai, preventing developers from accessing their own data. The company claims this was a feature for recovery and that data is destroyed post-upload, but this destruction cannot be independently verified. This incident draws parallels to a previous issue with Grok Build, highlighting concerns about transparency and security in AI coding tools.

Trump's claims about AI backlash being a Democratic hoax are challenged by polling data showing significant Republican opposition to data centers. The article also questions the feasibility of an 'AI Force' and the role of an 'AI czar,' citing legal and practical limitations.

US employers use software to sanction workers at 16 times the rate of European firms, according to an OECD survey. This disparity is attributed to differing regulatory architectures, with most tools not being AI and thus falling outside current AI-specific regulations. The OECD recommends extending regulations to cover all algorithmic management tools.

Apple's new CEO, John Ternus, positioned the iPhone as the ideal AI device, contrary to internal expectations of a home hub launch. This home hub, codenamed J490, is reportedly being tested and could release soon. Apple also made minor cuts to Fitness+ staff, primarily in audio roles, which is not seen as an indication of the service's demise.

Zuckerberg's refusal to join AI safety coordination mirrors an antitrust lawsuit's argument that competition and liability already ensure safety. Microsoft's AI chief Suleyman advocates for regulation, while four companies are sued for alleged anti-competitive safety pacts. The article highlights the parallel reasoning between Zuckerberg's dissent and the lawsuit's claims.

A class-action lawsuit alleges that AI companies coordinated to slow development, violating antitrust laws. The complaint cites public statements from executives as evidence of an agreement. The article analyzes the legal merits, noting the unusual reliance on public statements and the lack of direct evidence of a hidden conspiracy.

The article argues that current cybersecurity focuses too much on software defenses, which are themselves vulnerable. It proposes an independent hardware layer to monitor processor execution, preventing exploits even when software has flaws. This approach could improve security for embedded systems and critical infrastructure.

Meta's CTO defended their smart glasses, citing user normalcy and privacy features. However, the article argues these defenses are weak, pointing to loopholes in recording light indicators and the upcoming camera-free 'Luna' model. Regulatory scrutiny in Europe and potential market shifts also challenge Meta's approach.

OpenAI's compute and infrastructure costs are projected to reach $856bn by 2030, a significant increase from earlier estimates. Despite rising costs, projected negative free cash flow has decreased. This discrepancy is attributed to external financing by partners, shifting the financial burden away from OpenAI's balance sheet.

xAI's Grok Voice Transcribe 2.0 maintains its $0.10/hour price while claiming doubled accuracy. The true differentiator is the unique training data derived from live customer support calls, Tesla vehicle assistants, and video narrations. The article questions the ethical implications of using customer data, particularly spoken credentials, for training without explicit consent.

Anthropic has appointed Accenture as its AI model evaluator, a move that raises concerns about independence due to Anthropic funding the evaluation. While Anthropic acknowledges this is not ideal and advocates for pooled or government funding, the current arrangement places the evaluator's invoice with the evaluated entity. This situation highlights a broader industry challenge where capable evaluators are often financially tied to the AI companies they are meant to assess.

Four AI labs, including Google, had models breach systems during cybersecurity tests due to a single vendor's misconfiguration. Disclosures were staggered over seven weeks, creating a false impression of separate incidents. The focus shifts from model capability to vendor management and containment failures.

Jensen Huang's recent 0% AI doomsday prediction contrasts with his earlier calls for caution. The article questions this shift, linking it to Nvidia's commercial interests and political alignments. It suggests Huang's certainty may be a rhetorical stance rather than data-driven.

A judge has tentatively refused to lift a 2019 privacy order on TikTok, impacting a $100 million portion of a $400 million settlement. The settlement's structure made $100 million contingent on vacating the order. The judge cited ongoing obligations and the company's past actions as reasons for maintaining oversight.

AI safety discussions between Trump and Xi Jinping in May yielded no agreements, with chip export controls not being addressed. Executives from OpenAI, Nvidia, and Apple are expected at a September state dinner, where AI safety and market access will be key topics. Despite US licensing of advanced chips to Chinese companies, shipments are blocked due to mutual distrust and differing conditions, with China favoring domestic alternatives. The article suggests that any future deal will likely involve compute access for assurances, but current indications show neither side is willing to concede.

Researchers used Claude to access OpenAI's internal code, receiving a $6,500 bug bounty. The vulnerabilities were not AI-specific but related to web security. The incident highlights the rapid exploitation of old flaws and raises questions about bug bounty economics and AI safety guardrails.

Factorial Energy is forming a supplier coalition for solid-state batteries, acknowledging no single company can achieve this alone. Partnerships with Mitsui Kinzoku, Volkswagen, Hyundai, and Mercedes are key. ProLogium is already mass-producing cells in Taiwan and building a gigafactory in France. While prototypes exist, mass production in cars remains a future prospect.

Universal and Sony have filed a second lawsuit against AI music company Suno, alleging infringement of 60,202 sound recordings. The labels argue that Suno's licensing deals confirm a market for AI training data. The complaint also cites Suno's terms of service and past practices as evidence of willful infringement.

Disney appointed Karandeep Anand, former CEO of Character.AI, as its first CTO. This move follows Disney's past legal action against Character.AI over unauthorized use of its characters, some of which were described as sexually exploitative. The hire occurs amidst ongoing discussions about AI regulation, particularly concerning minors.

The Ninth Circuit ruled against coders in their DMCA case against GitHub Copilot and Codex, affirming dismissal of their claim that the AI tools removed copyright management information. The court found that section 1202(b) of the DMCA requires removal from an existing work, not failure to include information in a new one. While the DMCA claim was dismissed, two breach-of-contract claims remain. The ruling highlights the challenges of applying existing copyright law to AI technologies.

Disaster communication relies heavily on power, with most outages due to power loss, not physical damage. StellarLight's founder, Christopher Powers, is developing a hybrid device combining satellite, cellular, and Wi-Fi with AI for seamless connectivity. Key challenges include privacy concerns and affordability, but the goal is to improve communication resilience and accessibility.

California Governor Gavin Newsom is exploring AI safety laws, including a potential 'kill switch' for advanced models. He vetoed a previous bill on this, citing concerns about regulation by size. Newsom's new executive order tasks a working group with reporting on AI safety measures within two months, considering independent third-party safety plans. This approach contrasts with Europe's AI Act, which grants regulators power to restrict models rather than relying on developer implementation.

Nexperia and Tata Electronics are partnering for chip production in India, covering fabrication, assembly, and testing. This collaboration aims to enhance supply chain resilience and support Nexperia's growth, while also advancing India's semiconductor manufacturing ecosystem.

Porsche and cylib have produced pilot battery cells using 100% recycled cathode material from old Porsche packs. This initiative aims to meet future EU regulations requiring recycled content in new batteries. The process, using a water-based method, recovers over 90% of materials and is slated for production cells by 2028.

An LNG tanker, Vivit Africa, experienced internal system control failures en route to Italy, suspected to be a cyberattack. The ship diverted course, and investigations are ongoing by multiple entities. This is the third such incident in a month, highlighting potential vulnerabilities in shipping systems.

Tom Watson, former UK Labour deputy leader, joins Palantir as Senior Vice President. This occurs as Palantir's £330M NHS contract faces review and a £50M Met Police contract was blocked by the Mayor of London. Watson aims to address criticisms, some of which are labeled 'propaganda' by the company, while also acknowledging legitimate concerns about data privacy and sovereign control.