Musk OpenAI Trial Raises Trust Questions, Surprise AI Bills Shock Cloud Users

Tycoon2FA Shifts to Device Code Phishing to Hijack Microsoft 365 Accounts

• Tycoon2FA kit now abuses OAuth device authorization flows.
• Attackers obtain tokens via legitimate Microsoft login pages.
• Campaign persists after March 2026 coalition disruption.

USA, MAY 17 (TNGB) – Security researchers have documented how the Tycoon2FA phishing-as-a-service kit evolved after the March 2026 takedown by Microsoft, Europol and partners. The updated version lures targets into entering device codes on authentic Microsoft authentication pages, granting attackers session tokens for Microsoft 365 accounts without capturing passwords. The kit impersonates the Microsoft Authentication Broker application to facilitate token theft through the legitimate OAuth device code grant process.

Organizations face heightened risks from this refined identity attack vector.

Why This Matters: This shift forces defenders to monitor OAuth token grants more closely as traditional password and MFA controls prove insufficient against evolving kits.

Trust Questions Dominate Closing Arguments in Musk OpenAI Trial

• Federal trial in Oakland centers on OpenAI corporate structure.
• Testimony scrutinized Sam Altman credibility and founding commitments.
• Jury deliberations expected after arguments concluded mid-May.

OAKLAND, MAY 17 (TNGB) – Closing arguments wrapped in the high-profile federal trial between Elon Musk and OpenAI in Oakland, California. Musk’s legal team challenged the company’s shift from nonprofit roots to a for-profit model, while OpenAI countered that no binding charitable trust existed or was breached. The case has spotlighted internal communications and statements about AI development priorities and governance from the company’s early days.

The outcome could influence how courts view commitments made during AI company formations.

Why This Matters: A ruling on the charitable trust claims may set precedents for governance and mission adherence in rapidly scaling AI organizations.

Cloud Users Face Unexpected AI-Driven Billing Surges on Major Platforms

• Exposed API keys and heavy AI workloads triggered massive invoices.
• Google Cloud and AWS Bedrock users reported bills in tens of thousands.
• Anomaly detection tools sometimes failed to flag rapid spending.

USA, MAY 17 (TNGB) – Multiple cloud customers reported waking to unexpectedly large bills tied to AI services on Google Cloud and Amazon Bedrock. One documented case involved charges exceeding eighteen thousand dollars from a minimal budget after an exposed API key enabled thousands of requests. Another AWS user incurred over thirty thousand dollars despite configured cost anomaly detection thresholds. Providers have faced criticism for default settings and notification gaps during rapid AI experimentation.

Teams are now auditing API exposures and implementing stricter spending guardrails.

Why This Matters: Unchecked AI inference costs can quickly escalate operational expenses and strain budgets for organizations adopting generative tools.

Apple Explores Auto-Deleting Chats in Upcoming Siri Overhaul

• Planned standalone Siri app may add privacy-focused chat controls.
• Options reportedly include automatic deletion after set periods.
• Update aims to balance advanced capabilities with user data preferences.

USA, MAY 17 (TNGB) – Reports indicate Apple is preparing enhancements for Siri that could include a dedicated application experience with improved privacy options. Among the features under consideration is the ability for users to set automatic deletion of conversation history after thirty days or one year. The changes align with broader efforts to give users more granular control over data retention while potentially incorporating more capable underlying models.

Such adjustments reflect ongoing competition in voice assistant privacy features.

Why This Matters: Expanded deletion controls could influence user trust and adoption of AI assistants handling personal or sensitive queries.

DOJ Memo Outlines Federal Role in Voter Eligibility Data Collection

• Legal opinion asserts executive authority for nationwide vetting.
• Proposal draws on civil rights statutes for data access.
• State officials have labeled elements of the plan unrealistic.

WASHINGTON D.C., MAY 17 (TNGB) – A Department of Justice legal memorandum has asserted a robust federal role in collecting and reviewing voter eligibility data across states. The document frames the effort as consistent with existing civil rights enforcement authorities. Several state election officials have publicly questioned the feasibility and legal grounding of broad federal data demands, noting prior court skepticism toward similar initiatives.

Implementation would require coordination with state systems and face likely legal challenges.

Why This Matters: Expanded federal data collection efforts could reshape state-federal dynamics in election administration and verification processes.

Prusa Raises Alarms Over Bambu Lab Software Licensing and Security

• Josef Prusa alleges ongoing AGPL violations in Bambu slicer components.
• Closed-source network elements described as unauditable black boxes.
• Concerns include potential risks from foreign software in sensitive use cases.

USA, MAY 17 (TNGB) – 3D printing advocate Josef Prusa has renewed criticism of Bambu Lab over alleged violations of the AGPL license in its slicer software and related network plugins. Prusa and community developers point to closed-source components that handle network communication, calling them difficult to audit. Additional worries center on supply chain and data practices associated with the Chinese-origin hardware and software ecosystem, particularly for users in regulated industries.

The dispute has led to forks, legal threats, and community division.

Why This Matters: Licensing conflicts and security transparency issues in popular 3D printing tools can affect innovation, trust, and adoption in manufacturing and prototyping workflows.

Arizona Graduates Boo Eric Schmidt During AI-Focused Commencement Address

• Students interrupted former Google CEO at University of Arizona event.
• Boos followed comments on graduates shaping artificial intelligence.
• Reaction highlighted workforce concerns around automation and AI.

TUCSON, MAY 17 (TNGB) – Former Google CEO Eric Schmidt encountered audible boos from students and attendees while delivering the University of Arizona commencement address. The negative response intensified when Schmidt discussed the role graduates would play in developing and shaping artificial intelligence. Reports noted the disruption persisted through portions of the speech, reflecting broader anxieties about AI’s impact on entry-level jobs and career prospects.

The incident captured generational tensions around rapid technological change.

Why This Matters: Public pushback at university events signals growing student and public scrutiny of AI’s societal and employment effects.

OpenAI Launches Daybreak Initiative for AI-Assisted Cyber Defense

• Daybreak offers tiered access to GPT-5.5 models for security tasks.
• Platform targets vulnerability detection, patching, and validation.
• Effort positions OpenAI against more restricted competitor offerings.

USA, MAY 17 (TNGB) – OpenAI introduced Daybreak, a cybersecurity-focused program leveraging its advanced models including GPT-5.5 variants combined with agentic frameworks. The initiative provides different access tiers for defensive workflows such as secure code review, vulnerability triage, malware analysis, and patch validation. It contrasts with more tightly controlled alternatives by offering broader availability for authorized users while maintaining safeguards, with partners including major security firms.

Early adopters are exploring integration into existing security operations.

Why This Matters: AI tools purpose-built for cyber defense may accelerate vulnerability management but also raise questions about access controls and dual-use risks.

Classic 7 Mod Recreates Windows 7 Experience on Modern Windows 10 LTSC

• Community modification skins Windows 10 LTSC to mimic classic interface.
• Project appeals to users seeking nostalgia and legacy application compatibility.
• Base platform retains extended support through 2032.

USA, MAY 17 (TNGB) – Enthusiasts have released Classic 7, a modification that transforms the appearance and behavior of Windows 10 Long-Term Servicing Channel into a close visual and functional replica of Windows 7. The project preserves the security and stability benefits of the supported Windows 10 base while restoring familiar desktop elements and compatibility for older software. Users in gaming and specialized environments have shown particular interest.

Such community efforts demonstrate persistent demand for legacy user experiences.

Why This Matters: Nostalgia-driven modifications highlight gaps between modern operating system designs and user preferences for simplicity and familiarity.

Tokyo Researchers Unveil Ultrafast Low-Power Spintronic Switch for AI Hardware

• Antiferromagnetic device operates at picosecond speeds.
• Prototype generates minimal heat compared with conventional chips.
• Advance could enable more efficient next-generation computing architectures.

TOKYO, MAY 17 (TNGB) – Scientists at the University of Tokyo have demonstrated an antiferromagnetic spintronic switch capable of operating at picosecond timescales, roughly one thousand times faster than many current AI accelerator components. The Mn3Sn-based device reportedly produces very little waste heat, addressing key power and thermal constraints in high-performance computing. Researchers see potential applications in energy-efficient AI training and inference hardware.

Further engineering will determine commercial viability and integration paths.

Why This Matters: Fundamental hardware breakthroughs in speed and efficiency could reshape the economics and environmental footprint of large-scale AI systems.