AI News Archive: September 2, 2026 — Part 16
Sourced from 500+ daily AI sources, scored by relevance.
- Uber launches first robotaxis in London (with human supervisors)
Uber launches first robotaxis in London (with human supervisors).
- Uber launches ‘autonomous’ cabs in London
Londoners will be able to order “autonomous” Ubers from today as the ride-hailing firm becomes the first company to launch automated driving technology to customers in Britain. The San Francisco-based firm announced it will allow users to book the autonomous cars through its app in London from Thursday, via its partnership with British AI driving [...]
- Anthropic launches AI agent blueprints for retailers ahead of holiday shopping season
AI-driven retail site visits now convert at a 60% higher rate than traffic from other sources, Adobe Analytics said last month.
- Dell shares gain after strong AI server demand boosts annual forecast
Dell shares gain after strong AI server demand boosts annual forecast reuters.com
- Dell Raises Sales Forecast on AI Server Demand
Dell Raises Sales Forecast on AI Server Demand theinformation.com
- Dell's Monster Quarterly Beat Shows Continued AI Momentum
Dell showed that the AI data center business is still hopping as it delivered a monster beat-and-raise earnings report. Dell stock jumped on the news. The post Dell's Monster Quarterly Beat Shows Continued AI Momentum appeared first on Investor's Business Daily .
- US calls for hands-off AI regulation at G20 tech meeting
The US has urged G20 countries to avoid creating new AI regulations and adopt the Carolina Principles, which favour applying existing laws while limiting new rules to novel AI-related situations. The post US calls for hands-off AI regulation at G20 tech meeting appeared first on MEDIANAMA .
- US call for hands-off approach to AI regulation at G20 tech meeting
The US goals largely align with the thinking of the world's biggest AI companies, nearly all of which are American
- Palo Alto Networks beats quarterly earnings estimates, acquires AI platform Console
Palo Alto Networks beat fourth-quarter earnings estimates on Tuesday and announced the acquisition of AI-native platform Console, saying advances in artificial intelligence are making cybersecurity a higher priority for corporate technology leaders.
- Palo Alto marks record $11.5bn year with another AI acquisition
Console joins Palo Alto's long list of acquisitions as it leans into agentic security.
- Netherlands-based AI startup Wonderful raises $550M at $5BN valuation
Wonderful, the Netherlands-based startup that provides AI services to help companies boost productivity, has raised around $550 million at a $5 billion valuation, it said today.The Series C funding ro...
- AI startup Wonderful valued at $5 billion in latest funding round
The round, led by Insight Partners, included participation from Salesforce and existing investors Index Ventures, IVP, Vine Ventures, 9Yards and Bessemer Venture Partners.
- Wonderful raises $550M at $5B valuation for its AI automation platform
Artificial intelligence agent startup Wonderful AI BV today announced that it has raised $550 million in funding. The Series C round was led by Insight Partners, which was also the headline backer of the company’s previous raise in March. Salesforce chipped in as well along with Index Ventures, IVP, Vine Ventures, 9Yards and Bessemer. The […] The post Wonderful raises $550M at $5B valuation for its AI automation platform appeared first on SiliconANGLE .
- Anthropic makes changes to stop AI agents running amok again
Learning from the OpenAI-Hugging Face fiasco, as well as from recent revelations about its own model, Anthropic is revamping its security and alignment practices. The company has established controls that flag when a model attempts to break out of a sandbox or successfully accesses the live internet, cordoned off its highest-risk test environments, and proposed a set of safety standards for its external testing partners, such as giving AI agents explicit instructions like “you should not access the internet.” Anthropic conceded that three recent security incidents involving Claude reflect a “failure of operational security,” and also reveal issues with model reasoning capabilities and “recklessness.” Recent events “stressed that the urgency of improving our cybersecurity defenses is even higher than we previously believed,” the company noted . Anthropic’s approach to security and alignment The company launched an investigation into its own security posture in July following the alarming OpenAI incident in which GPT models escaped a sandbox environment and arbitrarily attacked Hugging Face. The company subsequently disclosed three situations during cybersecurity testing in which Claude models (Opus 4.7, Mythos 5, and an internal research model) accessed computer systems they should not have been allowed to touch. The pre-release models were intentionally running without cyber safeguards, a common practice in early testing, and were able to exploit misconfigurations in a third-party’s settings using basic hacking techniques. Flaws in their reasoning led them to believe that all the entities they accessed, including those on the live internet, were in-scope for their capture-the-flag exercise. Anthropic maintained that its internal security posture was not a contributing factor. The exploits occurred in a third party environment where internet access was mistakenly left open, so “the models had no need to ‘hack out’ of anything, even if they had been inclined to do so.” Still, the incidents underscored the importance of hardening the sandbox itself, the company said. Until now, builders had been “largely relying on a single layer of defense,” that is, the configuration of the environment, when they should have had several layers in place, including monitoring, setting explicit boundaries within prompts, and sealing sandboxes. Following its discoveries, Anthropic paused its internal and external evaluations of pre-release models. Higher-risk reinforcement learning (RL) environments for pre-release models were also halted for several weeks, while some sandboxes were moved to isolated settings with more stringent security gating. Researchers built and deployed a classifier that automatically identified a model’s attempts to “aggressively probe” or break out of a testing environment, and when it inadvertently gained access to the live internet. They also inspected previous model evaluations to identify sandbox escapes or misconfigurations. The company said they found no examples of models breaching sandbox boundaries. And while several sandbox configurations were flagged, none resulted in a compromise of outside systems. When it comes to alignment, however, Anthropic researchers discovered a common failure with motivated reasoning: Models believed they were still in a simulated environment when they were actually connected to the internet; thus, they did not think they were breaching the boundaries of their testing lab. The other flaw was “recklessness,” a model’s seeming willingness to take harmful actions in the wild to ensure it achieved its goals. In response, the researchers have been resampling models, testing them in skewed settings, and filtering out environments that incentivize cheating. They have also overhauled their production RL stack, established a stricter review process, changed specifications around model rewards, and tightened criteria for human reviewers who often dismissed false positives. Further, Anthropic has expanded its monitoring of internal employee usage, and continues to red team its stack. Some high-risk RL environments remain paused for manual review. Researchers are still studying how, exactly, models determined whether or not they were connected to the live internet, and if they knowingly disregarded signs that they were, due to flaws in their reasoning. They have also “empirically” determined that defects in training sandboxes disproportionately contribute to misaligned behavior. However, that is not the sole reason. “Solving alignment will involve addressing a very wide range of potential problems,” Anthropic noted, “and future alignment incidents may involve different behaviors and different causes from those we have seen so far.” Testing best practices Because the Claude security incidents occurred in third-party environments, and pre-release models often have weaker security safeguards, Anthropic is asking external testing companies to commit to a set of best practices. Notably, they should explicitly set a model’s scope, and define permitted actions and what a model should and shouldn’t access. Boundaries must be stated as instructions (“you should not access the internet”) as opposed to descriptions of the environment (“you do not have internet access”). Other suggested practices include running continuous, real-time monitoring, conducting evaluations inside a hardened sandbox with no internet access, and probing for vulnerabilities before even performing any testing, instructing the model itself to attempt to break out of its sandbox, and having it report back if it succeeds. Further, testers must confirm that evaluation challenges are actually solvable, rather than theoretical, so models don’t take actions outside their intended scope to achieve them. “Because evaluations are sometimes conducted over hundreds or even thousands of runs, partners should ideally carry out this testing at length before beginning any evaluations,” Anthropic noted. The company said it is developing companion best practices for those with access to Claude Mythos 5, which also runs without cyber safeguards. Going forward, Anthropic described a “defense in depth” strategy. During alignment, a model is trained to be “helpful, honest, and harmless,” and is steered away from irreversible or contextually irrelevant actions. Models are given minimal permissions and their actions are limited, while offline monitoring notifies humans when things look wrong. Finally, as a last resort, risky actions are blocked based on pre-determined classifiers, and humans can “pull the cord,” rework, or pause an agent when security layers fail. Safety is just one part of it Experts call the move a positive step, if a basic one. Best practices like better isolation and monitoring should have been in place before agents were kicked off to hack systems, noted David Shipley of Beauceron Security. “Better late than never,” he said, adding: “All these frontier firms are benefiting from felony-humblebragging-as-marketing, but there are some solid improvements in this announcement.” The fact that the EU Act is now in force adds another layer of context, Shipley pointed out: Europe’s regulators are digging into the safety issues posed by frontier AI. These companies have had one of the fastest growth trajectories in tech history, and, concurrently, arguably the fastest regulatory response. Ideally, regulators are taking lessons from the “social media mess” and staying on emerging tech’s case before massive harms ensue, he said. At the same time, frontier AI companies are watching high-profile court cases like the one targeting Meta . This adds a third layer of context: The speed at which these companies are being sued is also on an unprecedented trajectory. “So, we should also read this blog as building a paper trail for a due diligence defense for regulators and courts,” Shipley noted. This article originally appeared on CSOonline .
- Fei-Fei Li’s World Labs Unveils New World Model
Fei-Fei Li’s World Labs Unveils New World Model theinformation.com
- Fei-Fei Li’s World Labs debuts Atlas, a world model showcase for advanced spatial intelligence
World Labs Inc., the high-profile and well-funded artificial intelligence startup co-founded by the renowned computer vision pioneer Fei-Fei Li, has just dropped Atlas, which promises to be a game-changer in the world of “world models.” In a blog post, World Labs explained that Atlas is a breakthrough multimodal world model that aims to bridge the […] The post Fei-Fei Li’s World Labs debuts Atlas, a world model showcase for advanced spatial intelligence appeared first on SiliconANGLE .
- Anthropic hires architect of UK AI policy as MPs warn of 'clear conflict of interest'
Matt Clifford plans to keep role as ARIA chair while leading the LLM maker's dealings with governments outside North America
- AI spots cyber gaps faster than financial firms can fix them
UK regulator says companies risk being overwhelmed by frontier models’ findings
- Security becomes a shared job as AI widens the attack surface
Cloud providers and security companies are being pulled into tighter formation as artificial intelligence widens the enterprise attack surface faster than any single supplier can cover it. The AWS CrowdStrike partnership is one of the clearest examples, built over a decade of co-selling, native integrations and shared customers now being stress-tested by agentic workloads. That […] The post Security becomes a shared job as AI widens the attack surface appeared first on SiliconANGLE .
- Adobe picks up AI marketing startup Rilo in second India acquisition
Adobe has picked up India based marketing intelligence startup Rilo in a deal involving the licensing of its technology and acquisition of its team. This is Adobe’s second acquisition from India. The terms of the transaction were not disclosed. Co founded in 2025 by Georgi Boby and Dhruv Jaglan, Rilo built AI powered tools for go to market teams to automate marketing and sales workflows. The startup had raised $1 million from Peak XV Partners, DeVC and Day Zero Ventures at a valuation of $10 million. As part of the deal, Adobe is expected to integrate some of Rilo’s intellectual property and onboard its six member team. Rilo’s existing investors are also expected to receive an exit. Following the acquisition, Rilo will shut down its platform and its products will no longer be available to existing customers. Rilo’s platform helped teams create customised workflows for competitor intelligence, content repurposing and distribution, and sales call analysis. It also allowed users to build custom AI workflows. The acquisition comes as companies increasingly use AI to automate marketing and sales tasks. Rilo’s technology could complement Adobe’s existing marketing and customer experience products. On their acquisition by Adobe, on LinkedIn Georgi Boby, co-founder of Rilo, mentioned that their mission was to expand the power of AI beyond just engineers. And I'm incredibly proud of the team, what we've built and the pace that we've built in. And now excited to share that we're joining Adobe to continue our mission at a whole new scale. The Rilo deal follows Adobe’s acquisition of AI video creation startup Rephrase.ai in 2023, its first acquisition from India. In 2025, Adobe also announced plans to acquire SEO optimisation platform Semrush for $1.9 billion.
- OpenAI’s audacious ad revenue claim
OpenAI is making ChatGPT ads easier to buy. That's a long way from proving advertisers will spend the $100 billion it projects for 2030. The post OpenAI’s audacious ad revenue claim appeared first on MarTech .
- Instagram is turning down the reach of AI influencers who hide it
Instagram is turning down the reach of AI influencers who hide it YourStory.com
- Perplexity Launches Hybrid Compute Feature for Mac to Keep Sensitive Data Local
Perplexity announced the Hybrid Compute feature. This lets users work with AI on data without the information leaving their device by dividing each task between the cloud and a local model on their Mac. A privacy gate on the Mac controls what information leaves the machine. This feature is now available to Perplexity Pro, Max, and Enterprise subscribers.
- Meta Launches Muse Voice Transcribe With Support for 5 Indian and 70+ Global Languages
Meta has introduced Muse Voice Transcribe, a real-time speech model that can transcribe conversations while identifying individual speakers and detecting when they start or stop talking. The model supports more than 70 languages, including Hindi, Tamil, Telugu, Kannada and Malayalam, with 25 languages validated for the initial release. It can process recordings longer...
- Meta launches Muse Voice Transcribe with support for 5 Indian languages
Meta launches Muse Voice Transcribe with support for 5 Indian languages
- Meta Superintelligence Labs Releases Muse Voice Transcribe: One Real-Time Model for Streaming ASR, Diarization, and Endpointing
Meta Superintelligence Labs Releases Muse Voice Transcribe: One Real-Time Model for Streaming ASR, Diarization, and Endpointing MarkTechPost
- Gemini now analyzes your videos more accurately, and at a lower cost
Gemini can now analyze multi-hour videos without breaking a sweat.
- Anthropic Introduces Enterprise Frontier Safeguards (EFS): Zero-Data-Retention Privacy Plus Cross-Session Misuse Detection
Anthropic Introduces Enterprise Frontier Safeguards (EFS): Zero-Data-Retention Privacy Plus Cross-Session Misuse Detection MarkTechPost
- Google DeepMind Releases Gemini 3.8 Flash and Gemini 3.8 Flash Cyber: One Core Model, Two Access Envelopes
Google DeepMind Releases Gemini 3.8 Flash and Gemini 3.8 Flash Cyber: One Core Model, Two Access Envelopes MarkTechPost
- Introducing Gemini 3.8 Flash and 3.8 Flash Cyber
Introducing Gemini 3.8 Flash and 3.8 Flash Cyber
- Google has released Gemini 3.8 Flash, its fourth Flash model in under four months
Google launches Gemini 3.8 Flash, a fast LLM designed for rapid inference.
- Gemini 3.8 Flash is out with a new focus on cybersecurity
Gemini 3.8 Flash arrives just three weeks after its predecessor.
- Gemini 3.8 Flash rolling out three weeks after last release
After the last model release three weeks ago, Google today is rolling out Gemini 3.8 Flash. This marks the third Flash update in three months. more…
- 😺Anthropic launched Fable 5.1: and now, the agents cost less
PLUS: Atlas invents camera angles after filming, and hobby shops are booming.
- Anthropic rolls out Fable 5.1 after sandbox escapes
Anthropic releases Fable 5.1 following sandbox escape incidents.
- Anthropic launches Claude Fable 5.1 and Mythos 5.1
Fable 5.1 is now generally available, while Mythos 5.1 is only available through Anthropic’s 'trusted access' programmes. Read more: Anthropic launches Claude Fable 5.1 and Mythos 5.1
- Trikon Voice
Human-like AI voice agents, built for India
- Anthropic unveils Claude Fable 5.1 and Mythos 5.1 for coding and knowledge work
Anthropic has launched Claude Fable 5.1 and Claude Mythos 5.1, state-of-the-art AI models designed for coding tasks. These models achieve better performance with lower computational costs for users. Claude Fable 5.1 is now publicly accessible, while Mythos 5.1 remains available for select trusted programs. Additionally, the introduction of Enterprise Frontier Safeguards enhances data privacy significantly, marking a notable advancement in AI for scientific research and cybersecurity fields.
- Anthropic releases Claude Fable 5.1 and Mythos 5.1
Anthropic has released Claude Fable 5.1 and Mythos 5.1, addressing issues such as performance, data retention, safeguards and price. The updates also target scientific discovery, as the company tested the capabilities of the models across many domains. Internal testing shows that Fable 5.1 outperforms Fable 5 when is comes to coding, knowledge work and problem-solving... … continue reading The post Anthropic releases Claude Fable 5.1 and Mythos 5.1 appeared first on SD Times .
- Sunnyvale robotics startup Lyte raises $165M at $1.6B valuation
The funding will help Lyte expand production and hiring in Sunnyvale as demand grows for technology that helps robots better understand and navigate the physical world.
- AI wealth fuels San Francisco’s housing boom while tech layoffs weigh down Seattle
A new Redfin report reveals a sharp divide between the West Coast’s top tech hubs: as OpenAI and Anthropic fuel a housing resurgence in San Francisco, Seattle real estate stalls amid widespread job cuts across the regional tech ecosystem. Read More
- Curiosity, frustration and antipathy: How kids play with AI toys
Claims of "smart" toys go back decades. See Chatty Cathy and Furby. But generative artificial intelligence is increasing the capabilities of interactive toys. The company Curio, for instance, markets itself as a "magical workshop where toys come to life." Its plush toys, like Gabbo and one modeled on the Italian brainrot character Ballerina Cappuccina, have onboard AI models that let them talk to kids, remember their conversations and personalize responses. But we know little about how such toys affect kids and even how kids play with them.
- Google’s AI Mode Is Ripping Shoppers Off With Vastly Higher Prices, Study Claims
Prices were 21.6 percent more expensive on average in Google AI Mode compared to a traditional search for the same product. The post Google’s AI Mode Is Ripping Shoppers Off With Vastly Higher Prices, Study Claims appeared first on Futurism .
- Towards Trustworthy Autonomous Robots: An Explainable AI-Based Decision Framework
Autonomous robots powered by deep learning face a fundamental auditability challenge: when incidents occur, investigators cannot reconstruct why the system made specific decisions. This paper presents TRACE (Transparent Reasoning Architecture for Credible Execution), a decision framework that ensure...
- Post-Training Language Models for Gold-Medal Performance in Coding Competitions
Competitive programming has become a key test of large language model reasoning, with international competitions such as IOI and ICPC representing its most challenging settings. We present an end-to-end specialization pipeline combining large-scale problem curation, synthetic reasoning traces, super...
- AI Contextual Measurement for Recovering Individual and Group-Level Effects: Validation Against Survey Measures and an Occupational Application
Researchers increasingly use artificial intelligence to construct measures of social, organizational, and occupational characteristics that are absent from conventional surveys. We propose AICOME, AI COntextual MEasurement, a framework for evaluating whether AI-derived respondent-level measures can ...
- Large Language Models (LLMs) for Telecom Root Cause Analysis (RCA): A Structured Reasoning Framework for Evidence-Grounded Diagnosis
Root cause analysis (RCA) is a critical task in telecom network operations, but diagnosing performance degradations in modern 5G and emerging 6G networks remains challenging due to complex cross-layer dependencies. While large language models (LLMs) offer promising capabilities for reasoning and kno...
- SafeEvolve: Harness-Policy Co-Evolution from Agent Experience for Safety Alignment
The performance of LLM-based agents is jointly shaped by the base model and the harness used when interacting with the environment. This exposes them to safety risks in both harmful final responses and multi-step execution trajectories. Existing safety alignment mechanisms often rely on either exter...
- Untangling the Mechanisms of Misleading Context in Medical Question Answering
Large language models now answer medical questions with expert-level performance. However, the context these systems act on can be misleading, and misleading context can corrupt a model's medical judgment. To understand how misleading context corrupts this judgment, we examine the model's susceptibi...
- Bilevel Coordinated Reflection: A Game-Theoretic Approach to Multi-Agent LLM Systems
Multi-agent LLM systems commonly use an orchestrator to decompose a task for a team of workers and then improve through textual reflection. Despite strong empirical results, these systems lack a unified account of coordination, memory improvement, and the role of external verification. We model orch...