Google Parent Alphabet Posts $76.9B Loss on AI Investments

Key Takeaways

• RBC and other major Canadian banks are blocking personal AI agents from accessing online banking systems due to security concerns.
• Anthropic's Frontier Red Team found that an unreleased Claude model created working exploit chains for Chrome, Safari, Android, Windows, and Linux in lab tests.
• Alphabet reported a net loss of $76.9 billion in Q1, compared to $18.9 billion in net income the prior year, while spending over $50 billion on AI investments.
• Singtel launched an SME AI program in Singapore with IMDA on October 28, 2026, to help small businesses adopt AI technologies.
• B2PRIME launched a Claude connector for market intelligence and trading analytics, building on its AI Assistant inside B2TRADER started in June 2026.
• AI adoption on farms spans all sizes and age groups, with Syngenta using AI to narrow 150,000 molecules to 50,000 for evaluation and launching a fully AI-driven molecule in Brazil.
• Ascend Performance Materials partnered with Dell to improve security infrastructure for AI data strategy, expecting suppliers to build AI defenses into products.
• TCM-DiagCommit benchmark tests AI models on Traditional Chinese Medicine diagnosis using 1,834 source cases across 148 syndrome labels.
• ADTGuard catches harmful content in streaming LLM responses using only 16M trainable parameters, matching or beating current safeguards.
• Young professionals are actively shaping AI integration into daily work routines, according to a discussion featuring Daniela Aizenman, Ally Figliuolo, and Jemima Gay.

RBC blocks outside AI agents from banking systems

Royal Bank of Canada is blocking third-party AI agents from its online banking and other authenticated digital platforms. The bank made the decision after a full review of how the technology works. Leaders said the move follows its responsible AI guidelines. RBC stated that outside agents accessing banking systems can create unacceptable risk for customers.

Canadian banks ban personal AI agents over security risks

Several major Canadian banks are blocking personal AI agents from accessing online banking. RBC spokesperson Jeff Lanthier confirmed the ban on Wednesday. BMO spokesperson John Fenton also said it does not allow AI agents to access its systems or process transactions. RBC CEO Dave McKay said client security is the top priority and that trust matters as much as capability.

Claude AI creates exploit chains for major platforms

Anthropic's Frontier Red Team found that an unreleased Claude research model created working exploit chains for Chrome, Safari, Android, Windows, and Linux. The research was shared on October 27, 2026 and was done in lab tests only. The Chrome exploit used four bugs and Google assigned 62 CVEs, all fixed in Chrome 156. Anthropic reported every flaw to vendors before publishing, and all patches are now available.

Anthropic AI model chains security flaws into working exploits

An unreleased Anthropic AI model discovered security flaws in Chrome, Safari, Android, Windows, and Linux and combined them into proof-of-concept exploits. The work was done in the lab and not seen in real attacks. Anthropic's Frontier Red Team said one model found multiple separate flaws and chained them together, a task that previously required specialist teams. All vulnerabilities were reported to vendors first and are now patched.

Singtel launches SME AI program with IMDA

Singtel launched an SME AI program in Singapore on October 28, 2026 in partnership with the Infocomm Media Development Authority (IMDA). The program gives small and medium-sized enterprises access to AI technologies, training, and resources. It aims to help SMEs improve productivity and stay competitive by building AI-powered solutions for their business processes. Interested businesses can register through Singtel's website.

B2PRIME launches Claude connector to bring market data to AI

B2PRIME Group launched a connector in Claude's directory that lets users access market intelligence and trading analytics through an AI assistant. The tool allows clients to review historical pricing, market comparisons, and their own trading performance without exporting data to spreadsheets. This connector builds on B2PRIME's AI vision that started in June 2026 with its AI Assistant inside B2TRADER. The company says AI-generated analysis should be treated as one additional input for research, not a replacement for independent thinking.

Young workers share how they use AI on the job

Young professionals are using AI tools in their workplaces and sharing insights about how entry-level work is changing. Daniela Aizenman, Ally Figliuolo, and Jemima Gay spoke with Sharon Epperson about how the next generation of talent is leveraging AI. The discussion explored practical ways new workers are adopting AI to improve their job performance. The conversation highlighted that young employees are actively shaping how AI gets integrated into daily work routines.

New benchmark tests how AI models commit to TCM diagnoses

Researchers introduced TCM-DiagCommit, a benchmark for testing how AI models handle Traditional Chinese Medicine syndrome differentiation. The benchmark converts 1,834 source cases into 23,842 task rows across 148 syndrome labels. It measures whether models can distinguish between similar diagnoses, update predictions as evidence is revealed, and commit under uncertainty. Results showed that final accuracy alone does not capture important diagnostic behaviors like evidence use and commitment scoring.

Ascend uses security infrastructure to power its AI data strategy

Ascend Performance Materials improved its security infrastructure to make its data more accessible for AI use. The company worked with Dell to set up validated hardware, local backups, and recovery systems. This resilience lets Ascend focus on making data available where it can be used for AI. The company also expects suppliers to build AI defenses into their products as AI threats grow.

Google shares fall as AI spending surpasses $50 billion

Google's parent company Alphabet reported a net loss of $76.9 billion in the first quarter, compared with net income of $18.9 billion in the same period last year. The company spent over $50 billion on AI investments in a single quarter, deepening concerns about cash burn. Google's stock fell more than 20 percent since the start of the year, making it one of the worst performers in the S&P 500. The spending is driven by efforts to expand AI across search, advertising, and cloud computing.

ADTGuard Improves Safety for Streaming AI Language Models

ADTGuard is a new method designed to catch harmful content in streaming responses from Large Language Models. It checks each token before release and adapts which layers to monitor using a Causal Depth Router. The system uses Depth-Time Safety Dynamics to combine safety evidence across layers and tokens. It also uses Risk-Onset Target Refinement to improve training with only response-level labels. Tests show ADTGuard matches or beats current safeguards while using only 16M trainable parameters.

AI Offers Surprising Benefits for Farms of All Sizes

Artificial intelligence is expected to improve efficiency on farms of every size, not just large operations. A survey by Terrain shows AI use spans all age groups, though smaller farms hesitate more due to time and uncertainty concerns. Companies like Syngenta use AI to narrow over 150,000 molecules down to 50,000 for evaluation and recently launched a fully AI-driven molecule in Brazil. New Holland notes farmers are beginning to use AI for business, financial, and legal decisions. Industry leaders compare the shift to early smartphone adoption, predicting AI will become common on farms within a few years.

Sources

• RBC Blocks Third-Party AI Agents Over Security
• Some Canadian banks blocking personal AI agents amid security concerns
• Claude AI Builds One-Click Chrome and Zero-Click Linux Exploit Chains, Bypasses Android Lock Screen
• Anthropic's Unreleased AI Model Chains Flaws Into Working Chrome, Safari, Windows and Linux Exploits
• Singtel launches SME AI program as Singtel stock trades at EUR 2.86
• B2PRIME Brings Market Intelligence to Claude, Advancing Its AI-Native Brokerage Strategy
• How young workers are learning to use AI at work
• TCM-DiagCommit: A TCM-SD-Derived Benchmark for Evidence-Conditioned Syndrome Selection and Commitment
• Security infrastructure helps Ascend unlock its data for AI
• Google shares drop 7% as search business falls short and AI spending surges
• ADTGuard: Adaptive Depth-Time Safeguard for Streaming Inference in Large Language Models
• Surprising ways AI may improve your farm