Security experts warn "Black October" hacking surge will worsen before AI defenses catch up

Why "Black October" feels different from previous cyber incidents

Hari Mulackal, CEO of Canina Security, and Odysseas Lamtzidis, CEO of Filix, joined David Hoffman to discuss a wave of publicized security incidents. Mulackal said the underlying trend is not new: "the public incidents have been like very alarmist... the reality is that this has been going on for the last I would say 18 months." He traced a capability inflection point to around January, when AI models became cheap and effective enough to scan and exploit many companies at once.

Lamtzidis agreed, pointing to last year's deepfake-driven social engineering attacks as an earlier signal, now escalating into mass, automated scanning of the entire internet. He said "security through obscurity is breaking down in real time."

AI has turned hacking from a bespoke skill into a cheap, scalable factory

Mulackal argued that hacking used to require weeks or months of skilled human labor and a clear profit motive, making attacks against low-value targets uneconomical. Now, he said, "hacking has now become scalable where you can build a factory... it will cost like $10 as opposed to like two weeks of human time." He said this internal efficiency gain has made his team's own capabilities 15x cheaper since January, with further 10x drops expected every 6 to 12 months.

Lamtzidis added that this shift pulls attackers away from crypto specifically, since in traditional "web two" targets the monetization path used to be unclear. "Now that it's actually so cheap, you don't need to answer that question, you just do it and you'll figure it out at the end," he said.

Data exposure and extortion economics are the real-world harm

Mulackal described a shift toward extortion: attackers breach databases cheaply and then threaten companies with leaks unless paid. He cited CVE disclosure data as a symptom, saying a chart of cumulative vulnerability disclosures has gone "exponential" and that the cost of hacking is falling toward zero.

Both guests also warned about personalized AI-driven scams, citing an example where an AI engineer asked his personal AI assistant "tell me everything it knows about him" and received compensation details and information about romantic partners. Lamtzidis compared this to elder scams, now scalable to "millions of people, billions of people" with AI-generated believability.

"Software is getting built in factories today. What's going to happen in the next I would say two years would be security will also make a similar shift." — Hari Mulackal

Defense is lagging offense, but crypto may be better positioned

Mulackal estimated offense will dominate defense "for the next I would say 12 to 18 months," recommending companies attack themselves proactively using tools like his firm's product, Apex, to raise the cost of attacking them. He argued that making an attack cost $10,000 instead of $1 deters most opportunistic attackers.

Both guests suggested crypto firms, having evolved in an adversarial environment, are structurally ahead of traditional finance. Mulackal said Coinbase is "better prepared to defend" than a legacy bank like JP Morgan, though he stressed continued investment is still required. Hoffman offered this as a mildly bullish framing for crypto without the guests endorsing any price implication.

Practical advice: audit providers, diversify custody, avoid suspicious links

Mulackal urged listeners to audit the products and services holding their data or money, and to move away from providers who haven't visibly upgraded their security posture. Lamtzidis recommended building redundancy and limiting "blast radius," such as pairing a hardware wallet with a separate device so multiple supply chains would need to fail simultaneously for a loss to occur.

Lamtzidis also noted a market signal: he cited a 2024 Lloyd's statement that LLMs would increase cyber risk, and said cyber insurance premiums "have been falling consistently since 2024," predicting several insurance carriers may fail in coming months. On that basis, he suggested avoiding full self-custody for roughly the next 12 months in favor of spreading assets across exchanges with cyber insurance.

Where the guests disagreed or left things open

The guests gave differing timelines and framings for when things improve. Lamtzidis said after "the blood bath" things should return to "regular hand-to-hand combat" between attackers and defenders. Mulackal was more ambivalent, saying agent-driven defense will ultimately make software safer, but warned that a persistent AI attacker is "like an alien species that's now living among us," and that the world was never built for a one-mistake-and-you're-done environment. Neither gave a precise figure for how much data exposure or financial loss has already occurred, and both stopped short of saying current crypto holdings are either safe or unsafe in absolute terms.