Intellect-Partners

Categories
Computer Science

Ion Files New Patent to Stop Harmful AI Video Before It Ever Plays

Melbourne-based Ion is going after deepfakes and unsafe AI content with a system that blocks harmful video at the moment it’s built, not after it’s already online.

Most content moderation today works the same way: a video gets uploaded, it goes live, and only afterward does a platform check whether it breaks the rules. By the time anything gets flagged or removed, viewers have often already seen it.

Ion, an Australian Securities Exchange-listed company (ASX: IOV), wants to change that. The company has filed a new patent application in Australia for a system that checks video content for safety during assembly, before it’s ever displayed. If any piece fails a safety check, the video simply doesn’t get put together.

What the Patent Actually Covers

The filing is titled “System and Method for Content Safety Enforcement During Virtual Video Assembly.” Strip away the legal language, and the idea is fairly intuitive: instead of scanning a finished video file for problems, the system attaches safety rules to the individual building blocks of a video and checks those rules in real time as the video is constructed.

Ion calls this checking content at the “binary sample level.” In plain terms, the smallest underlying pieces of footage used to assemble a video each carry their own safety classification. When those pieces get pulled together into a final video, the system verifies every piece against the active rules before the output is allowed to exist.

If one piece fails, the video doesn’t resolve. There’s no finished file to flag or take down later, because the unsafe version was never created in the first place.

Why the Timing Matters

This approach is built for a specific shift happening across the video industry: content is increasingly generated or assembled on demand, rather than uploaded as a single static file.

Finbar O’Hanlon, Ion’s Chief Innovation Officer and the named inventor on the filing, framed the problem this way: moderation tools today were designed for finished files that can be reviewed before publication. But AI agents are starting to assemble video from fragments at the moment someone watches it, which often means there’s no complete file to inspect at all.

That’s the gap Ion is trying to close. The company describes its approach as binding a safety classification to every sample of content and enforcing it at the exact instant a video comes together, so that unsafe material never makes it to a screen in the first place.

Multiple Rulebooks, One Set of Source Files

One of the more practical aspects of the system is how it handles conflicting standards. Content rules can come from several places at once: the rights holder who owns the footage, the platform distributing it, and the jurisdiction where it’s being viewed. Local laws on AI-generated content, child safety, and disinformation vary widely from country to country, and platforms operating internationally have to satisfy all of them simultaneously.

Under Ion’s system, when those different rule sets disagree, the strictest standard automatically wins during assembly. That means a single set of source video files could be distributed across multiple regions, each governed by different legal requirements, without needing separate copies of the content built for each market.

For media companies and platforms managing global distribution, that’s a meaningful operational difference. It removes the need to maintain duplicate, region-specific versions of the same underlying content just to stay compliant.

Part of a Bigger IP Strategy

This isn’t Ion’s first patent in this space. The company previously filed for a system focused on video authentication, essentially a way to verify whether footage is genuine or AI-generated. This new filing addresses a related but distinct question: even if content is known to be real or fake, should it be allowed to be assembled and shown at all?

Together, the two filings cover different ends of the same problem. One asks whether a video is what it claims to be. The other asks whether it should be permitted to exist in its assembled form, regardless of its origin.

O’Hanlon positioned the newer patent as forward-looking, aimed at the kinds of problems platforms, broadcasters, studios, and regulators are increasingly being asked to solve as deepfakes, child safety risks, and disinformation campaigns become harder to police with traditional after-the-fact moderation.

Built to Expand Over Time

Ion says the system isn’t locked into a fixed set of rules. New categories of harm, new automated classifiers, and new regulatory requirements can be added through adapters without rebuilding the underlying media pipeline. That’s a notable design choice, since content regulation is a moving target. Laws governing AI-generated media are still being written in most major markets, and what counts as a prohibited category today may expand significantly within a few years.

By treating safety enforcement as a layer that sits alongside the core pipeline rather than something baked into it, Ion is positioning the system to adapt as both the technology and the legal landscape around it continue to shift.

The Bigger Picture

Whether or not this specific patent application is ultimately granted, it reflects a broader trend worth watching: as AI-generated and AI-assembled video becomes more common, the tools used to keep that content safe are having to move earlier in the pipeline. Reviewing a finished file after the fact works when there’s a finished file. It works much less well when video is being built dynamically, sample by sample, in response to a viewer.

Ion’s bet is that prevention at the point of assembly will matter more than detection after publication. For an industry racing to keep pace with generative AI, that’s a wager a lot of other companies may soon be making too.

Categories
Others

AI Patent Platform Fearn Secures $5.5M Seed Round to Automate Drafting

San Francisco-based intellectual property startup Fearn has announced the completion of a $5.5 million seed funding round to expand its AI-native patent drafting platform.

The round was led by Kindred Ventures, with participation from Andreessen Horowitz’s a16z speedrun startup accelerator, Designer Fund, and Essence VC. Prior to this institutional round, the company operated via founder self-funding.

The Founders and the Logic of Automation

Fearn was founded in 2025 by CEO Han Kim and CTO Angela Gao, who met as graduate students at Caltech. The platform’s architectural focus stems directly from the co-founders’ specialized backgrounds:

  • Han Kim: Previously prosecuted patent applications across software, life sciences, and mechanical arts as a scientific analyst at Morrison & Foerster, while researching bio-inspired neural algorithms during his Ph.D. track at Caltech.
  • Angela Gao: Completed a Ph.D. in computing and mathematical sciences at Caltech, specializing in physics-aligned generative models, alongside previous model development work at Google Research.

Kim noted that his experience in Big Law highlighted systemic inefficiencies in the traditional patent pipeline, which is frequently slow, cost-prohibitive, and anxiety-inducing for engineers worried about technical details being misinterpreted.

“I noticed a lot of the tasks I was doing could be automatable, but obviously I couldn’t automate them. You’re not really allowed to in those sorts of settings and environments,” Kim stated, highlighting the strict procedural friction within traditional law firms that inspired him to build an external automation solution.

How the Multi-Model Stack Works

Unlike general-purpose generative AI tools or simple API wrappers, Fearn is built from the ground up as a fully data-sovereign, AI-native platform. The coordinates a specialized multi-model stack:

  • Bespoke Model Ensemble: The platform utilizes dozens of hypercompact, specialized models, combining proprietary code, fine-tuned open-source models, and symbolic non-LLM systems built from scratch.
  • Data Sovereignty: Fearn hosts 100% of its own model stack internally. It makes zero application programming interface (API) calls to third-party model developers, completely removing the public-disclosure and data-egress risks that typically restrict enterprise IP teams from leveraging generative AI.
  • Hallucination Resistance: By training its custom architectures on highly curated, hand-corrected, and hand-labeled intellectual property datasets, Fearn creates audit trails engineered to guarantee compliance with patent office requirements and eliminate the factual errors common in large language models.

Once the application and automated labeled figures are ready, corporate research teams or solo inventors can choose to file the paperwork independently or hand it off to external counsel for final strategic review. Fearn charges a flat, predictable fee of $2,000 per patent draft, cutting traditional preparation timelines down significantly.

Future Plans and the Legal Tech Boom

With a lean team of fewer than 10 people, Fearn plans to deploy the capital injection primarily toward technical hiring, infrastructure expansion, and offsetting computational overhead.

Looking forward, the company intends to scale its features to assist inventors throughout the entire end-to-end patent prosecution lifestyle. This includes expanding automated systems to handle office action responses and any procedural workflow tied directly to a U.S. Patent and Trademark Office (USPTO) registration number.

Fearn’s successful seed round emphasizes an accelerating streak of legal tech investments by Andreessen Horowitz. The firm’s recent IP and legal portfolio expansion includes:

  • Leading patent automation startup Stilta’s seed round.
  • Anchoring multiple massive funding rounds for legal AI platform Harvey.
  • Backing litigation-focused developer Eve across two distinct rounds.
  • Leading the pre-seed round for communication security provider ZeroDrift.
Categories
Computer Science

Zero Trust Security: Why Traditional Firewalls Aren’t Enough

Introduction

With the quickly changing digital age, cybersecurity is now a top priority for organizations, governments, and individuals. Old security architectures that focus on perimeter defenses like firewalls are being left behind by the sophistication and magnitude of contemporary cyber threats. As organizations adopt cloud computing, remote work, and deploy Internet of Things (IoT) devices, the perimeter of a secure network is increasingly fuzzy. This new paradigm has spawned a new model of cybersecurity.

Zero Trust Security

The Legacy of Classic Firewalls

Fundamentally, a firewall is a software barrier that screens traffic through pre-programmed rules, separating what’s considered secure from what isn’t. Classically, firewalls have worked on the axiom that anything within a network can be trusted, and anything outside of it is potentially dangerous. This model was the foundation of enterprise security for many years.

The perimeter-based security architecture, though, was intended for a world in which everything users, applications, and data all sat behind an internal network. Employees typed away from office desktops, and sensitive data sat on on-premise servers. With these conditions, controlling access at the perimeter made sense. But the digital landscape has changed, rendering this model outdated and vulnerable.

The Limitations of Traditional Firewalls

The biggest failing of old firewalls is that it has in fact, eaten away at the network perimeter. The world has become cloud friendly, and is more inclined towards remote working with frequent access to the corporate resources from outside the office on personal devices or unsecured networks. Firewalls have minimal visibility or control over this activity. Therefore, attackers no longer must pierce strong network perimeters they just log in.

In addition, conventional firewalls are based on a model of implicit trust. Once a user or device gains once it enters the network, it is generally given wide access to internal systems and data. This creates an environment in which a single compromised endpoint can result in a catastrophic breach. Cyber attackers use this to their advantage by hijacking stolen credentials or malware to move laterally throughout the network, accessing sensitive information without popping up immediate red flags.

Another key problem is that firewalls are not cloud-native. Contemporary organizations tend to employ a hybrid of public cloud services, private data centers, and SaaS platforms. Firewalls, which were made for static environments, cannot enforce security policies uniformly in such dynamic infrastructures. Their rules and configurations are manually managed and hence are hard to scale and adapt in real-time.

What Is Zero Trust Security?

Zero Trust is an information technology framework based on the ideology of “never trust, always verify.” Unlike legacy models that trust that internal networks are secure, Zero Trust views every access request, whether it comes from within or without the organization, as suspicious. Access is
only permitted after rigorous identity authentication, device confirmation, and contextual risk evaluation.

In a Zero Trust architecture, least privilege access is a fundamental tenet. Users and devices are granted only the privileges they require to execute a particular set of tasks, nothing additional. This severely minimizes the attack surface and culls the potential impact if a credential is breached.

Micro-segmentation is another main characteristic of Zero Trust. Rather than depending on a solitary, integrated network, organizations segment their infrastructure into silos. Even should an attacker manage to get access into one segment of the system, they cannot easily move over to others. This resource compartmentalization achieves an additional layer of defence and constrains lateral movement.

Core Pillars of Zero Trust

A Zero Trust build consists of a number of inter-dependent elements:

  • Identity and Access Management (IAM): Verifies that only authenticated and approved users to access systems, usually with Multi-Factor Authentication (MFA) and Single Sign-On (SSO) for enhanced security.
  • Continuous Monitoring and Analytics: Zero Trust is not about static trust. The system continuously monitors user behaviour, location, device posture, and network activity to identify anomalies.
  • Device Trust: Not only is access granted based on user identity but also on the trustiness of the device. Is it patched? Is it in compliance with corporate policies?
  • Application and Data Security: Policies enforce secure access at the application layer, ensuring that users only interact with what’s required. Sensitive data is safeguarded through encryption, logging, and monitoring.
  • Zero Trust Network Access (ZTNA): ZTNA supplants traditional VPN solutions by linking users directly to particular applications instead of entire networks, thus reducing exposure.
Why Zero Trust Beats Firewalls

The contrast between Zero Trust and conventional firewalls isn’t philosophical it’s pragmatic.

Zero Trust presumes breach. It works with the expectation that attackers might already be within the network and constructs defences based on this expectation. Conventional firewalls are, on the other hand, reactive and concentrate on keeping threats external, too frequently neglecting what occurs once the perimeter is compromised.

Take the case of an attacker using a phishing attack to obtain valid user credentials. In the typical firewall-based setup, such an attacker would be able to penetrate the network and start exfiltrating information with minimal resistance. The firewall would not notice this internal traffic as malicious. But in a Zero Trust setup, the login attempt would initiate further verification processes. If the access is from an unexpected place or device, it might be blocked entirely. Even if the attacker successfully logs in, they would only have access to a thin slice of resources, and anomalous behaviour would most likely be picked up by analytics tools for rapid action.

Zero Trust in a Cloud-First World

The move towards cloud-native technologies and hybrid workspaces has made Zero Trust not only pertinent but necessary. Companies today are running on numerous cloud environments, SaaS offerings, and distributed teams geographically. Within these setups, the classical concept of “inside the network” is no longer applicable.

Zero Trust naturally belongs to this paradigm by taking security past the perimeter. It gives identity-based access control for all applications, services, and infrastructure irrespective of location. Whether a user is accessing from a corporate laptop in the office or a smartphone at home, their identity and behavior need to be authenticated prior to access.

Challenges of Implementing Zero Trust

Although it has its benefits, putting into practice Zero Trust is no cakewalk. For most organizations, it is a painstaking and resource-hungry process to move away from legacy systems. Identifying all applications, devices, users, and data flow across an organization is a serious task and is essential to effective Zero Trust adoption.

There is also a cultural side. Moving to Zero Trust can bring more restrictive access controls and increased authentication, which will likely meet opposition from users who are used to imperceptible access. With the right communications, training, and user experience design, though, these obstacles can be overcome.

In addition, vendors all don’t define Zero Trust similarly. Companies have diverse options for tools and platforms and need to exercise care in choosing them to verify they are aligned with real Zero Trust practices instead of marketing hype.

Conclusion:

Adopting the Cybersecurity Future
Historical firewalls were the initial defense line in a time when information was centralized, users were static, and threats were comparatively unsophisticated. But in today’s hyper-connected, decentralized digital space, these defenses are insufficient. The trust-based perimeter security model is inherently defective in a world where attacks can come from anywhere externally or internally.

Zero Trust Security provides a revolutionary strategy for this new world. By constantly verifying identity, implementing least privilege, and inspecting all attempts at access, organizations can construct strong, adaptive, and proactive security designs. Zero Trust is not a fad it’s a strategic imperative for any organization hoping to survive in the midst of today’s cyberattacks.

As cyberattacks increase in sophistication and the attack surface keeps growing, only those who trust no one and validate everything will stay safe.