Vol.11 · December 29, 2025
dera news AI Weekly Vol.11 | 2025-12-28 - This Week's AI News
🤖 dera news AI Weekly Vol.11
This week's AI world in one sentence?
As OpenAI reports reaching 1 million enterprise users for ChatGPT, a wave of copyright lawsuits against six major AI firms and the admission that prompt injection attacks are "unsolvable" by OpenAI itself underscore a critical reality check for the industry. This week, the narrative shifted from the boundless potential of AI to the urgent need for robust guardrails and careful operational strategies. The explosive corporate adoption of AI is now colliding head-on with escalating legal, security, and ethical friction.
📊 This Week's Question
Can AI's accelerating corporate adoption outrun its mounting legal, security, and ethical challenges?
This week presented a stark dichotomy: on one hand, undeniable evidence of AI's transformative power driving rapid business integration, and on the other, a surge in fundamental issues threatening to derail progress or at least complicate it significantly. It feels like we're simultaneously hitting the gas and discovering the brakes might be faulty.
Playing for "Acceleration & Adoption":
- Salesforce → Added 6,000 new enterprise customers in one quarter, driven by Agentforce's autonomous AI capabilities.
- EdgeVerve → Showcased Agentic AI's ability to autonomously evolve and manage complex enterprise tasks.
- Lovable → Secured $330M at a $6.6B valuation, proving strong investor confidence in AI-powered no-code development.
- Liquid AI → Released LFM2-2.6B-Exp, a smaller, smarter model expanding efficient AI capabilities to edge devices.
- Waymo/Google → Began testing Gemini-powered AI assistants in robotaxis, enhancing autonomous service with conversational AI.
Playing for "Friction & Reality Check":
- Six major AI firms (including Anthropic) → Hit with new copyright lawsuits over alleged use of pirated books in training data.
- OpenAI → Acknowledged that prompt injection attacks are "not fully solvable," highlighting a persistent security vulnerability.
- Italian Authorities → Issued a temporary halt to Meta's policy prohibiting rival AI chatbots on WhatsApp, citing antitrust concerns.
- Indie Game Awards → Revoked a Game of the Year award from "Clair Obscur: Expedition 33" due to undisclosed AI usage in assets.
What we're watching closely is OpenAI's candid admission that prompt injection attacks are "not fully solvable."
This is noteworthy because it's a rare moment of technical humility from a leading AI developer, signaling that some security challenges might be inherent to current LLM architectures. It forces us to rethink our defensive strategies, moving beyond "fixing" to "mitigating" and "monitoring."
This week's takeaway: The era of simply deploying AI tools for novelty is over. Businesses must now navigate a complex landscape where rapid innovation demands equally rapid, and robust, risk assessment and compliance. You can't win with just one anymore; you need both agility and strong guardrails.
💡 This Week's Actions
1. Review your AI data sourcing and usage policies for copyright compliance. (1-2 hours) Given the fresh wave of copyright lawsuits against major AI firms, it's critical to understand the provenance of data your AI models are trained on or that your AI tools interact with. Even if you're using third-party models, understanding their data policies and ensuring your own input data is clean can mitigate future legal risks. → Understanding AI Copyright Issues
2. Audit your prompt engineering practices for prompt injection vulnerabilities. (3-4 hours) OpenAI's admission that prompt injection is "not fully solvable" means you can't assume your AI applications are immune. Review your critical internal and external-facing AI prompts and workflows. Implement defensive prompt design (e.g., clear delimiters, negative constraints) and robust output validation to minimize exposure. → OpenAI's Acknowledgment of Prompt Injection
3. Identify one small, repetitive business workflow ripe for agentic AI automation. (2-3 hours) The rise of agentic AI and platforms like Lovable demonstrates that autonomous agents are becoming more accessible and powerful. Pick a low-risk, high-volume task (e.g., data aggregation, report generation from structured inputs) and explore how a simple agent could handle it, starting with a proof-of-concept. This allows you to learn the benefits and risks of "hiring" AI without deep technical expertise. → Explore Agentic AI's Potential
💡 For Individual Practitioners
Paying $20/month for ChatGPT Plus but not seeing the results you expected?
We've compiled 50 practical use cases for business professionals who want to master ChatGPT.
👉 ChatGPT Practical Guide - 50 Use Cases
From sales materials to data analysis and meeting notes — concrete examples you can use tomorrow.
📰 This Week's AI Articles (All 9)
1️⃣ Six Major AI Firms Face New Copyright Lawsuits
🏷️ Topic: IP & Regulation
What Happened? Six prominent AI companies, including Anthropic, Stability AI, and Midjourney, have been hit with a fresh wave of copyright infringement lawsuits by authors and publishers. The allegations center on the unauthorized use of copyrighted books, including potentially pirated copies, as training data for their large language models (LLMs) and generative AI art tools. This legal offensive signals a significant escalation in the ongoing battle over intellectual property rights in the age of generative AI.
Our take What's notable here isn't just the lawsuits themselves, but the collective target and the specific allegations of using pirated content. This suggests a growing legal strategy to challenge the foundational data practices of leading AI developers. For businesses leveraging AI, this signals an increased need for due diligence on the provenance of AI models and a clear understanding of their own data input strategies to avoid becoming collateral damage.
2️⃣ Salesforce Sees 6,000 New Enterprise Customers in Three Months, Driven by Agentic AI
🏷️ Topic: Strategic Partnership & Business Adoption
What Happened? Salesforce reported a significant surge in its enterprise customer base, adding 6,000 new clients in the last quarter, marking a 48% increase. A key driver behind this growth is their Agentforce platform, an autonomous AI agent solution, which saw its customer count jump from 12,500 to 18,500. This indicates a strong market appetite for AI solutions that move beyond simple chatbots to more autonomous, workflow-integrating agents within the enterprise.
Our take This massive customer growth for Agentforce is a clear signal that businesses aren't just experimenting with AI anymore; they're actively integrating and scaling autonomous solutions. What we're watching is how Salesforce's success will push other enterprise software giants to accelerate their own agentic AI offerings, potentially creating a new competitive frontier beyond basic LLM integrations. It solidifies the shift from "AI assistance" to "AI automation."
3️⃣ OpenAI Admits "Prompt Injection" Attacks Are "Not Fully Solvable"
🏷️ Topic: Ethics & Safety
What Happened? In a security report concerning its new "ChatGPT Atlas" feature, OpenAI openly acknowledged that prompt injection attacks, where malicious inputs manipulate an AI model's behavior, are "not fully solvable." This admission highlights an inherent and persistent vulnerability in current large language models, suggesting that complete immunity to such attacks may be technically unachievable. The report underscores the expanding "attack surface" as AI agents become more autonomous and integrated into business workflows.
Our take This is a critical "reality check" from the industry leader. What's notable is the directness of OpenAI's statement, forcing a fundamental shift in how we approach AI security – from trying to eliminate vulnerabilities to focusing on robust mitigation and monitoring strategies. We read this as a clear warning that organizations deploying AI agents must prioritize layered defenses and continuous oversight, rather than relying on the AI provider for a magic bullet solution.
4️⃣ Agentic AI Transforms Enterprise Operations Beyond Traditional AI Assistance
🏷️ Topic: LLM Dev & Business Adoption
What Happened? EdgeVerve is championing Agentic AI, a paradigm shift from traditional AI that merely assists humans to AI that autonomously evolves and drives business processes. Unlike conventional AI tools that wait for human commands, Agentic AI is designed to anticipate challenges, integrate diverse systems, and proactively manage complex workflows. This evolution enables businesses to automate entire segments of operations, moving towards truly intelligent automation.
Our take This signals a maturation of AI from a "copilot" to a "navigator" and "driver" for business processes. What's notable is the focus on prediction and integration, which moves beyond simple task execution to genuine problem-solving. We see this as a key trend for SMEs looking to achieve significant efficiency gains, but it also amplifies the need for careful oversight and robust data governance given the increased autonomy.
5️⃣ Italy Orders Meta to Halt WhatsApp AI Chatbot Policy Amid Regulatory Scrutiny
🏷️ Topic: Regulation & Ethics & Safety
What Happened? Italian authorities have issued a temporary injunction against Meta, ordering the company to suspend its policy that prohibits rival AI chatbots on WhatsApp. This regulatory intervention stems from concerns over potential anti-competitive practices and the stifling of innovation within the AI ecosystem. The move highlights growing scrutiny from global regulators regarding how dominant platforms manage and integrate AI services, especially concerning third-party developers.
Our take This signals a proactive stance from regulators to ensure a level playing field in the rapidly evolving AI landscape. What's notable is the focus on preventing major platforms from leveraging their market position to dictate AI usage, which could significantly impact how small and medium-sized businesses can integrate diverse AI solutions. This is a reminder that regulatory friction will continue to be a significant factor in AI deployment strategies.
6️⃣ AI Coding Platform Lovable Reaches $6.6 Billion Valuation After $330 Million Series B Round
🏷️ Topic: LLM Dev & Strategic Partnership
What Happened? Swedish AI coding startup Lovable has achieved a staggering $6.6 billion valuation after successfully raising $330 million in its Series B funding round. The company's platform allows users to create applications and websites from simple text descriptions, eliminating the need for traditional programming. This no-code approach, powered by advanced AI, is democratizing software development and attracting significant investor confidence.
Our take Lovable's meteoric rise underscores the immense market demand for tools that lower the barrier to entry for software creation. What's notable is its potential to empower SMEs and non-technical founders to rapidly prototype and deploy digital solutions, bypassing traditional development bottlenecks. This signals a future where "coding" becomes more about clear communication with AI than writing lines of syntax, which is a massive opportunity for efficiency.
7️⃣ Indie Game Awards Revokes "Game of the Year" Title Over Undisclosed AI Usage
🏷️ Topic: Ethics & Safety & IP
What Happened? "Clair Obscur: Expedition 33," a game that had won the prestigious Game of the Year award at the Indie Game Awards, has had its title revoked. The unprecedented decision came after it was revealed that the game utilized generative AI for certain assets without clear disclosure. This incident has ignited a fierce debate within the creative community about transparency, authenticity, and the ethical implications of AI integration in artistic endeavors.
Our take This is a stark example of the "content ecosystem crisis" emerging from generative AI. What's notable is that the issue isn't necessarily AI usage itself, but the lack of transparency, highlighting a growing tension between creators, platforms, and AI tools regarding ethical disclosure. We read this as a clear message that authenticity and clear communication about AI's role will be crucial for maintaining trust and credibility in creative industries.
8️⃣ Liquid AI Unveils Smarter, Smaller Models for Enhanced Edge AI Efficiency
🏷️ Topic: LLM Dev & Hardware
What Happened? Liquid AI has introduced LFM2-2.6B-Exp, a new small-scale language model designed for enhanced performance in instruction following, knowledge tasks, and mathematics. Despite its compact size, which allows it to run efficiently on edge devices, the model demonstrates significantly improved capabilities. This development promises substantial benefits for SMEs by offering robust AI functionalities with reduced computational costs and enhanced data security, as processing can occur locally.
Our take This is a game-changer for democratizing advanced AI, especially for smaller businesses. What's notable is the combination of improved intelligence with a smaller footprint, addressing key concerns around cost, privacy, and latency. We see this as a strong signal that powerful AI is becoming less reliant on massive cloud infrastructure, opening up new possibilities for secure, on-device applications across various industries.
9️⃣ Waymo Tests Gemini-Powered AI Assistant in Robotaxis to Enhance Passenger Experience
🏷️ Topic: LLM Dev & Strategic Partnership
What Happened? Waymo, Google's autonomous driving subsidiary, has begun testing an in-cabin AI assistant powered by Google's Gemini large language model within its robotaxi fleet. This development, uncovered through an analysis of system prompts, reveals the assistant's ability to answer general knowledge questions and control vehicle features like lighting and temperature. The integration aims to enrich the passenger experience by providing conversational interaction and personalized controls.
Our take This signals a fascinating convergence of autonomous driving and conversational AI, moving beyond just getting from A to B to enhancing the entire journey. What's notable is how Gemini's capabilities are being leveraged to create a more intuitive and comfortable passenger environment, showcasing the potential for AI to personalize and humanize even highly automated services. We're watching how this integration will set new standards for in-vehicle user experience.
📚 Editor's Note
This week, the news that truly made our editorial team pause was the confluence of escalating copyright lawsuits and OpenAI's admission about prompt injection. It’s a sobering reminder that while AI's capabilities are expanding at an incredible pace, the fundamental challenges of intellectual property, security, and ethical boundaries are becoming more urgent, not less.
Why we consider it important is that these aren't peripheral issues; they strike at the core of AI's trustworthiness and long-term viability. It forces us all to move beyond the initial "wow" factor of generative AI and engage with the complex realities of deploying it responsibly.
At dera news, we believe this is precisely why our role is so crucial. We're not just here to report on how "AI is amazing," but to create a space where we can think together about "how to use AI" effectively, ethically, and securely in the real world. This week, more than ever, felt like a collective reality check, and we're here to help you navigate it.
We'll be back next week with useful information and food for thought.
dera news Editorial Team
🤝 Ready for Team-Wide AI Adoption?
Once you've tested ChatGPT individually and felt its potential, the next step is team-wide implementation.
But where do you start? Which tools to choose? How to roll out internally? How to measure ROI?
Let's clarify these challenges with our AI adoption assessment.
We'll discuss your business needs and challenges, then propose the optimal implementation roadmap. After the assessment, we can introduce our full AI Team implementation support if needed.
📬 About this newsletter